remove using namespace useless

Tensorrt8

.gitignore

@@ -1,10 +1,4 @@
 *~
-demo/demo/data/img_crop/
-demo/demo/data/img_disparity/
-demo/demo/data/map/
-demo/demo/data/masks_orient/
-demo/demo/data/pmat_new/
-demo/demo/data/masks_v2/
 build/
 .vscode/
 *.bin
@@ -14,6 +8,17 @@ build/
 *.h5
 *.tar.gz
 *.weights
-*.zip
 .idea/
 *.hdf5
+*.pk
+*.table
+cmake-build-release/
+demo/COCO_val2017
+demo/BDD100K_val
+/.vs
+cmake-build-minsizerel/*
+scripts/COCO_val2017/*
+scripts/COCO_val2017.zip
+scripts/all_labels.txt
+/cmake/cuda_script
+/cmake-build-debug/

.gitmodules

@@ -1,6 +0,0 @@
-[submodule "tracker_CLASS"]
-	path = tracker_CLASS
-	url = https://github.com/mive93/tracker_CLASS.git
-[submodule "masa_protocol"]
-	path = masa_protocol
-	url = https://git.hipert.unimore.it/rcavicchioli/masa_protocol.git

CMakeLists.txt

@@ -1,8 +1,69 @@
-cmake_minimum_required(VERSION 3.5)
-
-project (tkDNN)
+cmake_minimum_required(VERSION 3.15)
+project(tkDNN)
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fPIC")
+set(CMAKE_CXX_STANDARD 14)
+
+option(ENABLE_OPENCV_CUDA_CONTRIB "Enable OpenCV CUDA Contrib" OFF )
+
+if(NOT CMAKE_BUILD_TYPE)
+  set(CMAKE_BUILD_TYPE "Release" CACHE STRING "default build" FORCE)
+endif(NOT CMAKE_BUILD_TYPE)
+
+find_package(CUDA 9.0 REQUIRED)
+if (CUDA_FOUND)
+    set(OUTPUTFILE ${CMAKE_CURRENT_SOURCE_DIR}/cmake/cuda_script) # No suffix required
+    execute_process(COMMAND "rm ${OUTPUTFILE}")
+    set(CUDAFILE ${CMAKE_CURRENT_SOURCE_DIR}/cmake/getCudaArch.cu)
+    execute_process(COMMAND ${CUDA_NVCC_EXECUTABLE} -lcuda ${CUDAFILE} -o ${OUTPUTFILE})
+    execute_process(COMMAND ${OUTPUTFILE}
+            RESULT_VARIABLE CUDA_RETURN_CODE
+            OUTPUT_VARIABLE ARCH)
+
+    if(${CUDA_RETURN_CODE} EQUAL 0)
+        set(CUDA_SUCCESS "TRUE")
+    else()
+        set(CUDA_SUCCESS "FALSE")
+    endif()
+
+    if (${CUDA_SUCCESS})
+        message(STATUS "CUDA Architecture: ${ARCH}")
+        message(STATUS "CUDA Version: ${CUDA_VERSION_STRING}")
+        message(STATUS "CUDA Path: ${CUDA_TOOLKIT_ROOT_DIR}")
+        message(STATUS "CUDA Libararies: ${CUDA_LIBRARIES}")
+        message(STATUS "CUDA Performance Primitives: ${CUDA_npp_LIBRARY}")
+        set(CUDA_NVCC_FLAGS "${ARCH}")
+    else()
+        message(WARNING ${ARCH})
+    endif()
+endif()
+
+SET(CUDA_SEPARABLE_COMPILATION ON)
+
+if(UNIX)
+    if(CMAKE_BUILD_TYPE MATCHES Release)
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}  -fPIC -Wno-deprecated-declarations -Wno-unused-variable -O3")
+        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32)
+    endif()
+
+    if(CMAKE_BUILD_TYPE MATCHES Debug)
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}  -fPIC -Wno-deprecated-declarations -Wno-unused-variable -g3")
+        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32 -G -g)
+    endif()
+endif()
+
+if(WIN32)
+    if(CMAKE_BUILD_TYPE MATCHES Release)
+        set(CMAKE_CXX_FLAGS "/O2 /FS /EHsc /MD")
+        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32)
+    endif()
+
+    if(CMAKE_BUILD_TYPE MATCHES Debug)
+        set(CMAKE_CXX_FLAGS "/Od /FS /EHsc /MDd")
+        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32 -G -g)
+    endif()
+set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
+endif(WIN32)
+
 include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include/tkDNN)
 
 # project specific flags
@@ -10,54 +71,77 @@ if(DEBUG)
     add_definitions(-DDEBUG)
 endif()
 
+if(TKDNN_PATH)
+    message("SET TKDNN_PATH:" ${TKDNN_PATH})
+    add_definitions(-DTKDNN_PATH="${TKDNN_PATH}")
+else()
+    add_definitions(-DTKDNN_PATH="${CMAKE_CURRENT_SOURCE_DIR}")
+endif()
+
 
 #-------------------------------------------------------------------------------
 # CUDA
 #-------------------------------------------------------------------------------
-find_package(CUDA 9.0 REQUIRED)
-SET(CUDA_SEPARABLE_COMPILATION ON)
-#set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -arch=sm_30 --compiler-options '-fPIC'")
+set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}" --compiler-options '-fPIC')
+
 
 find_package(CUDNN REQUIRED)
+include_directories(${CUDNN_INCLUDE_DIR})
+
+find_package(yaml-cpp REQUIRED)
+
 
 # compile
-file(GLOB tkdnn_CUSRC "src/kernels/*.cu")
+file(GLOB tkdnn_CUSRC "src/kernels/*.cu" "src/sorting.cu" "src/pluginsRT/*.cpp")
 cuda_include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include ${CUDA_INCLUDE_DIRS} ${CUDNN_INCLUDE_DIRS})
 cuda_add_library(kernels SHARED ${tkdnn_CUSRC})
+target_link_libraries(kernels ${CUDA_CUBLAS_LIBRARIES} ${CUDA_LIBRARIES} ${CUDNN_LIBRARIES} yaml-cpp)
+
 
 
 #-------------------------------------------------------------------------------
 # External Libraries
 #-------------------------------------------------------------------------------
+find_package(Eigen3 REQUIRED)
+message("Eigen DIR: " ${EIGEN3_INCLUDE_DIR})
+include_directories(${EIGEN3_INCLUDE_DIR})
+
 find_package(OpenCV REQUIRED)
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DOPENCV")
+if(ENABLE_OPENCV_CUDA_CONTRIB)
+    if (OpenCV_FOUND)
+        find_package(OpenCV COMPONENTS cudawarping cudaarithm)
+        if(OpenCV_cudawarping_FOUND AND OpenCV_cudaarithm_FOUND)
+            add_compile_definitions(OPENCV_CUDACONTRIB)
+            message("OpenCV Cuda Contrib modules found")
+        else()
+            message("OpenCV Cuda Contrib modules not found")
+            set(ENABLE_OPENCV_CUDA_CONTRIB OFF)
+        endif()
+    endif()
+endif()
+# if(OpenCV_CUDA_VERSION)
+#     add_compile_definitions(OPENCV_CUDACONTRIB)
+# endif()
 
-include_directories(/usr/include/gdal)
+# gives problems in cross-compiling, probably malformed cmake config
 
 #-------------------------------------------------------------------------------
 # Build Libraries
 #-------------------------------------------------------------------------------
 file(GLOB tkdnn_SRC "src/*.cpp")
-set(tkdnn_LIBS kernels ${CUDA_LIBRARIES} ${CUDA_CUBLAS_LIBRARIES} ${CUDNN_LIBRARIES} ${OpenCV_LIBS})
-
-file(GLOB class_SRC "src/class_src/*.cpp")
-
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -std=c++11 -O3")
-include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include ${CUDA_INCLUDE_DIRS} ${OPENCV_INCLUDE_DIRS} ${NVINFER_INCLUDES} "~/repos/cereal/include" ${CMAKE_CURRENT_SOURCE_DIR}/tracker_CLASS/c++/src /usr/include/python2.7)
-
-set(class_LIBS ${OpenCV_LIBS} -lgdal yaml-cpp python2.7)
+set(tkdnn_LIBS kernels ${CUDA_LIBRARIES} ${CUDA_CUBLAS_LIBRARIES} ${CUDNN_LIBRARIES} ${OpenCV_LIBS} yaml-cpp)
 
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include ${CUDA_INCLUDE_DIRS} ${OPENCV_INCLUDE_DIRS} ${NVINFER_INCLUDES})
 add_library(tkDNN SHARED ${tkdnn_SRC})
-target_link_libraries(tkDNN ${tkdnn_LIBS})
-
-add_library(CLASS SHARED ${class_SRC})
-target_link_libraries(CLASS ${class_LIBS})
-
+target_link_libraries(tkDNN ${tkdnn_LIBS} ${CUDA_CUBLAS_LIBRARIES})
 
 #static 
 #add_library(tkDNN_static STATIC ${tkdnn_SRC})
 #target_link_libraries(tkDNN_static ${tkdnn_LIBS})
 
+# SMALL NETS
 add_executable(test_simple tests/simple/test_simple.cpp)
 target_link_libraries(test_simple tkDNN)
 
@@ -67,62 +151,93 @@ target_link_libraries(test_mnist tkDNN)
 add_executable(test_mnistRT tests/mnist/test_mnistRT.cpp)
 target_link_libraries(test_mnistRT tkDNN)
 
-## YOLO NETS
-add_executable(test_yolo tests/yolo/yolo.cpp)
-target_link_libraries(test_yolo tkDNN)
-
-add_executable(test_yolo_voc tests/yolo_voc/yolo_voc.cpp)
-target_link_libraries(test_yolo_voc tkDNN)
-
-add_executable(test_yolo_tiny tests/yolo_tiny/yolo_tiny.cpp)
-target_link_libraries(test_yolo_tiny tkDNN)
-
-add_executable(test_yolo_relu tests/yolo_relu/yolo_relu.cpp)
-target_link_libraries(test_yolo_relu tkDNN)
-
-
-add_executable(test_yolo_224 tests/yolo_224/yolo_224.cpp)
-target_link_libraries(test_yolo_224 tkDNN)
-
-add_executable(test_yolo_berkeley tests/yolo_berkeley/yolo_berkeley.cpp)
-target_link_libraries(test_yolo_berkeley tkDNN)
-
-add_executable(test_yolo3_coco4 tests/yolo3_coco4/yolo3_coco4.cpp)
-target_link_libraries(test_yolo3_coco4 tkDNN)
-
-add_executable(test_yolo3_berkeley tests/yolo3_berkeley/yolo3_berkeley.cpp)
-target_link_libraries(test_yolo3_berkeley tkDNN)
-
-add_executable(test_yolo3_tetrapack tests/yolo3_tetrapack/yolo3_tetrapack.cpp)
-target_link_libraries(test_yolo3_tetrapack tkDNN)
-
-add_executable(test_yolo3_tetrapack_resize tests/yolo3_tetrapack_resize/yolo3_tetrapack_resize.cpp)
-target_link_libraries(test_yolo3_tetrapack_resize tkDNN)
-
-add_executable(test_yolo3_BCDS6 tests/yolo3_BCDS6/yolo3_BCDS6.cpp)
-target_link_libraries(test_yolo3_BCDS6 tkDNN)
-add_executable(test_yolo3_flir tests/yolo3_flir/yolo3_flir.cpp)
-target_link_libraries(test_yolo3_flir tkDNN)
-
 add_executable(test_imuodom tests/imuodom/imuodom.cpp)
 target_link_libraries(test_imuodom tkDNN)
-################################################################################
+
+# DARKNET
+file(GLOB darknet_SRC "tests/darknet/*.cpp")
+foreach(test_SRC ${darknet_SRC})
+    get_filename_component(test_NAME "${test_SRC}" NAME_WE)
+    set(test_NAME test_${test_NAME})
+    add_executable(${test_NAME} ${test_SRC})
+    target_link_libraries(${test_NAME} tkDNN)
+    install(TARGETS ${test_NAME} DESTINATION bin)
+endforeach()
+
+# MOBILENET
+add_executable(test_mobilenetv2ssd tests/mobilenet/mobilenetv2ssd/mobilenetv2ssd.cpp)
+target_link_libraries(test_mobilenetv2ssd tkDNN)
+
+add_executable(test_bdd-mobilenetv2ssd tests/mobilenet/bdd-mobilenetv2ssd/bdd-mobilenetv2ssd.cpp)
+target_link_libraries(test_bdd-mobilenetv2ssd tkDNN)
+
+add_executable(test_mobilenetv2ssd512 tests/mobilenet/mobilenetv2ssd512/mobilenetv2ssd512.cpp)
+target_link_libraries(test_mobilenetv2ssd512 tkDNN)
+
+# BACKBONES
+add_executable(test_resnet101 tests/backbones/resnet101/resnet101.cpp)
+target_link_libraries(test_resnet101 tkDNN)
+
+add_executable(test_dla34 tests/backbones/dla34/dla34.cpp)
+target_link_libraries(test_dla34 tkDNN)
+
+# CENTERNET
+add_executable(test_resnet101_cnet tests/centernet/resnet101_cnet/resnet101_cnet.cpp)
+target_link_libraries(test_resnet101_cnet tkDNN)
+
+add_executable(test_dla34_cnet tests/centernet/dla34_cnet/dla34_cnet.cpp)
+target_link_libraries(test_dla34_cnet tkDNN)
+
+add_executable(test_dla34_cnet3d tests/centernet/dla34_cnet3d/dla34_cnet3d.cpp)
+target_link_libraries(test_dla34_cnet3d tkDNN)
+
+# CENTERTRACK
+
+add_executable(test_dla34_ctrack tests/centertrack/dla34_ctrack/dla34_ctrack.cpp)
+target_link_libraries(test_dla34_ctrack tkDNN)
+
+# SHELFNET
+add_executable(test_shelfnet tests/shelfnet/shelfnet.cpp)
+target_link_libraries(test_shelfnet tkDNN)
+
+add_executable(test_shelfnet_berkeley tests/shelfnet/shelfnet_berkeley.cpp)
+target_link_libraries(test_shelfnet_berkeley tkDNN)
+
+add_executable(test_shelfnet_mapillary tests/shelfnet/shelfnet_mapillary.cpp)
+target_link_libraries(test_shelfnet_mapillary tkDNN)
+
+add_executable(test_shelfnet_coco tests/shelfnet/shelfnet_coco.cpp)
+target_link_libraries(test_shelfnet_coco tkDNN)
+
+# MONODEPTH2
+add_executable(test_monodepth2_640 tests/monodepth2/monodepth2_640.cpp)
+target_link_libraries(test_monodepth2_640 tkDNN)
+
+add_executable(test_monodepth2_1024 tests/monodepth2/monodepth2_1024.cpp)
+target_link_libraries(test_monodepth2_1024 tkDNN)
 
 
+# DEMOS
 add_executable(test_rtinference tests/test_rtinference/rtinference.cpp)
 target_link_libraries(test_rtinference tkDNN)
 
-add_executable(yolo3_demo demo/demo/demo.cpp 
-                        tracker_CLASS/c++/src/ekf.cpp
-                        tracker_CLASS/c++/src/trackutils.cpp
-                        tracker_CLASS/c++/src/plot.cpp
-                        tracker_CLASS/c++/src/tracker.cpp )
+add_executable(map_demo demo/demo/map.cpp)
+target_link_libraries(map_demo tkDNN)
 
-                        
-target_link_libraries(yolo3_demo tkDNN CLASS)
+add_executable(demo demo/demo/demo.cpp)
+target_link_libraries(demo tkDNN)
 
+add_executable(demo3D demo/demo/demo3D.cpp)
+target_link_libraries(demo3D tkDNN)
 
+add_executable(demoTracker demo/demo/demoTracker.cpp)
+target_link_libraries(demoTracker tkDNN)
 
+add_executable(seg_demo demo/demo/seg_demo.cpp)
+target_link_libraries(seg_demo tkDNN)
+
+add_executable(demoDepth demo/demo/demoDepth.cpp)
+target_link_libraries(demoDepth tkDNN)
 
 #-------------------------------------------------------------------------------
 # Install
@@ -133,23 +248,11 @@ target_link_libraries(yolo3_demo tkDNN CLASS)
 #endif()
 message("install dir:" ${CMAKE_INSTALL_PREFIX})
 install(DIRECTORY include/ DESTINATION include/)
-install(TARGETS tkDNN kernels DESTINATION lib)
+install(TARGETS tkDNN  DESTINATION lib)
+install(TARGETS test_simple test_mnist test_mnistRT test_rtinference demo map_demo DESTINATION bin)
 install(DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/cmake/" # source directory
         DESTINATION "share/tkDNN/cmake/" # target directory
 )
-
-
-#-------------------------------------------------------------------------------
-# Prepare for test
-#-------------------------------------------------------------------------------
-set(TEST_DATA true CACHE BOOL "If true download deps")
-if( ${TEST_DATA} )
-	message("Launching pre-build dependency installer script...")
-
-    execute_process (COMMAND bash -c "bash build_models.sh download"
-        WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/tests)
-
-	set(TEST_DATA false CACHE BOOL "If true download deps" FORCE)
-	message("Finished dowloading test weights")
-endif()
-
+install(DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/tests/" # source directory
+        DESTINATION "share/tkDNN/tests" # target directory
+)

LICENSE

@@ -0,0 +1,339 @@
+                    GNU GENERAL PUBLIC LICENSE
+                       Version 2, June 1991
+
+ Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
+ 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The licenses for most software are designed to take away your
+freedom to share and change it.  By contrast, the GNU General Public
+License is intended to guarantee your freedom to share and change free
+software--to make sure the software is free for all its users.  This
+General Public License applies to most of the Free Software
+Foundation's software and to any other program whose authors commit to
+using it.  (Some other Free Software Foundation software is covered by
+the GNU Lesser General Public License instead.)  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+this service if you wish), that you receive source code or can get it
+if you want it, that you can change the software or use pieces of it
+in new free programs; and that you know you can do these things.
+
+  To protect your rights, we need to make restrictions that forbid
+anyone to deny you these rights or to ask you to surrender the rights.
+These restrictions translate to certain responsibilities for you if you
+distribute copies of the software, or if you modify it.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must show them these terms so they know their
+rights.
+
+  We protect your rights with two steps: (1) copyright the software, and
+(2) offer you this license which gives you legal permission to copy,
+distribute and/or modify the software.
+
+  Also, for each author's protection and ours, we want to make certain
+that everyone understands that there is no warranty for this free
+software.  If the software is modified by someone else and passed on, we
+want its recipients to know that what they have is not the original, so
+that any problems introduced by others will not reflect on the original
+authors' reputations.
+
+  Finally, any free program is threatened constantly by software
+patents.  We wish to avoid the danger that redistributors of a free
+program will individually obtain patent licenses, in effect making the
+program proprietary.  To prevent this, we have made it clear that any
+patent must be licensed for everyone's free use or not licensed at all.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                    GNU GENERAL PUBLIC LICENSE
+   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+  0. This License applies to any program or other work which contains
+a notice placed by the copyright holder saying it may be distributed
+under the terms of this General Public License.  The "Program", below,
+refers to any such program or work, and a "work based on the Program"
+means either the Program or any derivative work under copyright law:
+that is to say, a work containing the Program or a portion of it,
+either verbatim or with modifications and/or translated into another
+language.  (Hereinafter, translation is included without limitation in
+the term "modification".)  Each licensee is addressed as "you".
+
+Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope.  The act of
+running the Program is not restricted, and the output from the Program
+is covered only if its contents constitute a work based on the
+Program (independent of having been made by running the Program).
+Whether that is true depends on what the Program does.
+
+  1. You may copy and distribute verbatim copies of the Program's
+source code as you receive it, in any medium, provided that you
+conspicuously and appropriately publish on each copy an appropriate
+copyright notice and disclaimer of warranty; keep intact all the
+notices that refer to this License and to the absence of any warranty;
+and give any other recipients of the Program a copy of this License
+along with the Program.
+
+You may charge a fee for the physical act of transferring a copy, and
+you may at your option offer warranty protection in exchange for a fee.
+
+  2. You may modify your copy or copies of the Program or any portion
+of it, thus forming a work based on the Program, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+    a) You must cause the modified files to carry prominent notices
+    stating that you changed the files and the date of any change.
+
+    b) You must cause any work that you distribute or publish, that in
+    whole or in part contains or is derived from the Program or any
+    part thereof, to be licensed as a whole at no charge to all third
+    parties under the terms of this License.
+
+    c) If the modified program normally reads commands interactively
+    when run, you must cause it, when started running for such
+    interactive use in the most ordinary way, to print or display an
+    announcement including an appropriate copyright notice and a
+    notice that there is no warranty (or else, saying that you provide
+    a warranty) and that users may redistribute the program under
+    these conditions, and telling the user how to view a copy of this
+    License.  (Exception: if the Program itself is interactive but
+    does not normally print such an announcement, your work based on
+    the Program is not required to print an announcement.)
+
+These requirements apply to the modified work as a whole.  If
+identifiable sections of that work are not derived from the Program,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works.  But when you
+distribute the same sections as part of a whole which is a work based
+on the Program, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Program.
+
+In addition, mere aggregation of another work not based on the Program
+with the Program (or with a work based on the Program) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+  3. You may copy and distribute the Program (or a work based on it,
+under Section 2) in object code or executable form under the terms of
+Sections 1 and 2 above provided that you also do one of the following:
+
+    a) Accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of Sections
+    1 and 2 above on a medium customarily used for software interchange; or,
+
+    b) Accompany it with a written offer, valid for at least three
+    years, to give any third party, for a charge no more than your
+    cost of physically performing source distribution, a complete
+    machine-readable copy of the corresponding source code, to be
+    distributed under the terms of Sections 1 and 2 above on a medium
+    customarily used for software interchange; or,
+
+    c) Accompany it with the information you received as to the offer
+    to distribute corresponding source code.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form with such
+    an offer, in accord with Subsection b above.)
+
+The source code for a work means the preferred form of the work for
+making modifications to it.  For an executable work, complete source
+code means all the source code for all modules it contains, plus any
+associated interface definition files, plus the scripts used to
+control compilation and installation of the executable.  However, as a
+special exception, the source code distributed need not include
+anything that is normally distributed (in either source or binary
+form) with the major components (compiler, kernel, and so on) of the
+operating system on which the executable runs, unless that component
+itself accompanies the executable.
+
+If distribution of executable or object code is made by offering
+access to copy from a designated place, then offering equivalent
+access to copy the source code from the same place counts as
+distribution of the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+  4. You may not copy, modify, sublicense, or distribute the Program
+except as expressly provided under this License.  Any attempt
+otherwise to copy, modify, sublicense or distribute the Program is
+void, and will automatically terminate your rights under this License.
+However, parties who have received copies, or rights, from you under
+this License will not have their licenses terminated so long as such
+parties remain in full compliance.
+
+  5. You are not required to accept this License, since you have not
+signed it.  However, nothing else grants you permission to modify or
+distribute the Program or its derivative works.  These actions are
+prohibited by law if you do not accept this License.  Therefore, by
+modifying or distributing the Program (or any work based on the
+Program), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Program or works based on it.
+
+  6. Each time you redistribute the Program (or any work based on the
+Program), the recipient automatically receives a license from the
+original licensor to copy, distribute or modify the Program subject to
+these terms and conditions.  You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties to
+this License.
+
+  7. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Program at all.  For example, if a patent
+license would not permit royalty-free redistribution of the Program by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Program.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system, which is
+implemented by public license practices.  Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+  8. If the distribution and/or use of the Program is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Program under this License
+may add an explicit geographical distribution limitation excluding
+those countries, so that distribution is permitted only in or among
+countries not thus excluded.  In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+  9. The Free Software Foundation may publish revised and/or new versions
+of the General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+Each version is given a distinguishing version number.  If the Program
+specifies a version number of this License which applies to it and "any
+later version", you have the option of following the terms and conditions
+either of that version or of any later version published by the Free
+Software Foundation.  If the Program does not specify a version number of
+this License, you may choose any version ever published by the Free Software
+Foundation.
+
+  10. If you wish to incorporate parts of the Program into other free
+programs whose distribution conditions are different, write to the author
+to ask for permission.  For software which is copyrighted by the Free
+Software Foundation, write to the Free Software Foundation; we sometimes
+make exceptions for this.  Our decision will be guided by the two goals
+of preserving the free status of all derivatives of our free software and
+of promoting the sharing and reuse of software generally.
+
+                            NO WARRANTY
+
+  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
+FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
+OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
+PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
+TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
+PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
+REPAIR OR CORRECTION.
+
+  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
+REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
+INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
+OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
+TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
+YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
+PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGES.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+convey the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    tkDNN 
+    Copyright (C) 2017 Francesco Gatti
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+Also add information on how to contact you by electronic and paper mail.
+
+If the program is interactive, make it output a short notice like this
+when it starts in an interactive mode:
+
+    Gnomovision version 69, Copyright (C) year name of author
+    Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, the commands you use may
+be called something other than `show w' and `show c'; they could even be
+mouse-clicks or menu items--whatever suits your program.
+
+You should also get your employer (if you work as a programmer) or your
+school, if any, to sign a "copyright disclaimer" for the program, if
+necessary.  Here is a sample; alter the names:
+
+  Yoyodyne, Inc., hereby disclaims all copyright interest in the program
+  `Gnomovision' (which makes passes at compilers) written by James Hacker.
+
+  <signature of Ty Coon>, 1 April 1989
+  Ty Coon, President of Vice
+
+This General Public License does not permit incorporating your program into
+proprietary programs.  If your program is a subroutine library, you may
+consider it more useful to permit linking proprietary applications with the
+library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.

README.md

@@ -1,58 +1,215 @@
 # tkDNN
-tkDNN is a Deep Neural Network library built with cuDNN primitives specifically thought to work on NVIDIA TK1(and all successive) board.<br>
-The main scope is to do high performance inference on already trained models.
+tkDNN is a Deep Neural Network library built with cuDNN and tensorRT primitives, specifically thought to work on NVIDIA Jetson Boards. It has been tested on TK1(branch cudnn2), TX1, TX2, AGX Xavier, Nano and several discrete GPUs.
+The main goal of this project is to exploit NVIDIA boards as much as possible to obtain the best inference performance. It does not allow training. 
 
-this branch actually work on every NVIDIA GPU that support the dependencies:
-* CUDA 10.0
-* CUDNN 7.603
-* TENSORRT 6.01
-* OPENCV 4.1
 
-## Dependencies
+If you use tkDNN in your research, please cite the [following paper](https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9212130&casa_token=sQTJXi7tJNoAAAAA:BguH9xCIY48MxbtDS3LXzIXzO-9sWArm7Hd7y7BwaLmqRuM_Gx8bOYizFPNMNtpo5K0kB-P-). For use in commercial solutions, write at gattifrancesco@hotmail.it and micaela.verucchi@unimore.it or refer to https://hipert.unimore.it/ .
 
 ```
-sudo apt install libgdal-dev libeigen3-dev python-matplotlib libyaml-cpp-dev libcereal-dev python2.7-dev
+@inproceedings{verucchi2020systematic,
+  title={A Systematic Assessment of Embedded Neural Networks for Object Detection},
+  author={Verucchi, Micaela and Brilli, Gianluca and Sapienza, Davide and Verasani, Mattia and Arena, Marco and Gatti, Francesco and Capotondi, Alessandro and Cavicchioli, Roberto and Bertogna, Marko and Solieri, Marco},
+  booktitle={2020 25th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA)},
+  volume={1},
+  pages={937--944},
+  year={2020},
+  organization={IEEE}
+}
+```
+
+### What's new
+#### 20 July 2021
+- [x] Support to sematic segmentation [README](docs/README_seg.md)
+- [x] Support 2D/3D Object Detection and Tracking [README](docs/README_2d3dtracking.md)
+#### 24 November 2021
+- [x] Support to sematic segmentation on cuda 11
+- [x] Support to TensorRT8. (thanks to [Harshvardhan Chandirasekar](https://github.com/perseusdg))
+#### 30 March 2022
+- [x] Support to monocular depth esitmation [README](docs/README_depth.md) (thanks to [Harshvardhan Chandirasekar](https://github.com/perseusdg))
+
+
+
+## FPS Results
+Inference FPS of yolov4 with tkDNN, average of 1200 images with the same dimension as the input size, on 
+  * RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5);
+  * Xavier AGX, Jetpack 4.3 (CUDA 10.0, CUDNN 7.6.3, tensorrt 6.0.1 );
+  * Xavier NX, Jetpack 4.4  (CUDA 10.2, CUDNN 8.0.0, tensorrt 7.1.0 ). 
+  * Tx2, Jetpack 4.2 (CUDA 10.0, CUDNN 7.3.1, tensorrt 5.0.6 );
+  * Jetson Nano, Jetpack 4.4  (CUDA 10.2, CUDNN 8.0.0, tensorrt 7.1.0 ). 
+
+| Platform   | Network    | FP32, B=1 | FP32, B=4	| FP16, B=1 |	FP16, B=4 |	INT8, B=1 |	INT8, B=4 | 
+| :------:   | :-----:    | :-----:   | :-----:   | :-----:   |	:-----:   |	:-----:   |	:-----:   | 
+| RTX 2080Ti | yolo4 320  | 118.59	  | 237.31	  | 207.81	  | 443.32	  | 262.37	  | 530.93    | 
+| RTX 2080Ti | yolo4 416  | 104.81	  | 162.86	  | 169.06	  | 293.78	  | 206.93	  | 353.26    | 
+| RTX 2080Ti | yolo4 512  | 92.98	    | 132.43	  | 140.36	  | 215.17	  | 165.35	  | 254.96    | 
+| RTX 2080Ti | yolo4 608  | 63.77	    | 81.53	    | 111.39	  | 152.89	  | 127.79	  | 184.72    | 
+| AGX Xavier | yolo4 320  |	26.78	    | 32.05	    | 57.14	    | 79.05	    | 73.15	    | 97.56     |
+| AGX Xavier | yolo4 416  |	19.96	    | 21.52	    | 41.01	    | 49.00	    | 50.81	    | 60.61     |
+| AGX Xavier | yolo4 512  |	16.58	    | 16.98	    | 31.12	    | 33.84	    | 37.82	    | 41.28     |
+| AGX Xavier | yolo4 608  |	9.45 	    | 10.13	    | 21.92	    | 23.36	    | 27.05	    | 28.93     |
+| Xavier NX  | yolo4 320  |	14.56	    | 16.25	    | 30.14	    | 41.15	    | 42.13	    | 53.42     |
+| Xavier NX  | yolo4 416  |	10.02	    | 10.60	    | 22.43	    | 25.59	    | 29.08	    | 32.94     |
+| Xavier NX  | yolo4 512  |	8.10	    | 8.32	    | 15.78	    | 17.13	    | 20.51	    | 22.46     |
+| Xavier NX  | yolo4 608  |	5.26	    | 5.18	    | 11.54	    | 12.06	    | 15.09	    | 15.82     |
+| Tx2        | yolo4 320	| 11.18	    | 12.07	    | 15.32	    | 16.31     | -         | -         |
+| Tx2        | yolo4 416	| 7.30	    | 7.58	    | 9.45	    | 9.90      | -         | -         |
+| Tx2        | yolo4 512	| 5.96	    | 5.95	    | 7.22	    | 7.23      | -         | -         |
+| Tx2        | yolo4 608	| 3.63	    | 3.65	    | 4.67	    | 4.70      | -         | -         |
+| Nano       | yolo4 320	| 4.23	    | 4.55	    | 6.14	    | 6.53      | -         | -         |
+| Nano       | yolo4 416	| 2.88	    | 3.00	    | 3.90	    | 4.04      | -         | -         |
+| Nano       | yolo4 512	| 2.32	    | 2.34	    | 3.02	    | 3.04      | -         | -         |
+| Nano       | yolo4 608	| 1.40	    | 1.41	    | 1.92	    | 1.93      | -         | -         |
+
+## MAP Results
+Results for COCO val 2017 (5k images), on RTX 2080Ti, with conf threshold=0.001
+
+|                      | CodaLab       | CodaLab   | CodaLab       | CodaLab     | tkDNN map     | tkDNN map |
+| -------------------- | :-----------: | :-------: | :-----------: | :---------: | :-----------: | :-------: |
+|                      | **tkDNN**     | **tkDNN** | **darknet**   | **darknet** | **tkDNN**     | **tkDNN** |
+|                      | MAP(0.5:0.95) | AP50      | MAP(0.5:0.95) | AP50        | MAP(0.5:0.95) | AP50      |
+| Yolov3 (416x416)     | 0.381         | 0.675     | 0.380         | 0.675       | 0.372         | 0.663     |
+| yolov4 (416x416)     | 0.468         | 0.705     | 0.471         | 0.710       | 0.459         | 0.695     |
+| yolov3tiny (416x416) | 0.096         | 0.202     | 0.096         | 0.201       | 0.093         | 0.198     |
+| yolov4tiny (416x416) | 0.202         | 0.400     | 0.201         | 0.400       | 0.197         | 0.395     |
+| Cnet-dla34 (512x512) | 0.366         | 0.543     | \-            | \-          | 0.361         | 0.535     |
+| mv2SSD (512x512)     | 0.226         | 0.381     | \-            | \-          | 0.223         | 0.378     |
+
+## Index
+- [tkDNN](#tkdnn)
+  - [Index](#index)
+  - [Dependencies](#dependencies)
+  - [How to compile this repo](#how-to-compile-this-repo)
+  - [Workflow](#workflow)
+  - [Exporting weights](#exporting-weights)
+  - [Run the demos](#run-the-demos)
+  - [tkDNN on Windows 10 or Windows 11](#tkdnn-on-windows-10-or-windows-11)
+  - [Existing tests and supported networks](#existing-tests-and-supported-networks)
+  - [References](#references)
+  
+
+## Dependencies
+This branch works on every NVIDIA GPU that supports the following (latest tested) dependencies:
+* CUDA 11.3 (or >= 10.2) 
+* cuDNN 8.2.1 (or >= 8.0.4)
+* TensorRT 8.0.3 (or >=7.2)
+* OpenCV 4.5.4 (or >=4)
+* cmake 3.21 (or >= 3.15)
+* yaml-cpp 0.5.2
+* eigen3 3.3.4
+* curl 7.58
+
+```
+sudo apt install libyaml-cpp-dev curl libeigen3-dev
+
+```
+
+#### About OpenCV
+To compile and install OpenCV4 with contrib us the script ```install_OpenCV4.sh```. It will download and compile OpenCV in Download folder.
+```
+bash scripts/install_OpenCV4.sh
+```
+If you have OpenCV compiled with cuda and contrib and want to use it with tkDNN pass  ```ENABLE_OPENCV_CUDA_CONTRIB=ON``` flag when compiling tkDBB
+. If the flag is not passed,the preprocessing of the networks is computed on the CPU, otherwise on the GPU. In the latter case some milliseconds are saved in the end-to-end latency. 
+
+## How to compile this repo
+Build with cmake. If using Ubuntu 18.04 a new version of cmake is needed (3.15 or above).
+On both linux and windows ,the ```CMAKE_BUILD_TYPE``` variable needs to be defined as either ```Release``` or ```Debug```.
+```
+git clone https://github.com/ceccocats/tkDNN
+cd tkDNN
+mkdir build
+cd build
+cmake -DCMAKE_BUILD_TYPE=Release .. 
+make
 ```
 
 ## Workflow
-The recommended workflow follow these step:
-* Build and train a model in Keras (on any PC)
-* Export weights and bias 
-* Define the model on tkDNN
-* Do inference (on TK1)
+Steps needed to do inference on tkDNN with a custom neural network. 
+* Build and train a NN model with your favorite framework.
+* Export weights and bias for each layer and save them in a binary file (one for layer).
+* Export outputs for each layer and save them in a binary file (one for layer).
+* Create a new test and define the network, layer by layer using the weights extracted and the output to check the results. 
+* Do inference.
 
-## Compile the library
-Build with cmake
-```
-mkdir build
-cd build
-cmake ..
-# use -DTEST_DATA=False to skip dataset download
-make
-```
-during the cmake configuration it will be dowloaded the weights needed for running
-the tests
+## Exporting weights
 
-## Test
-Assumiung you have correctly builded the library these are the test ready to exec:
-* test_simple: a simple convolutional and dense network (CUDNN only)
-* test_mnist: the famous mnist netwok (CUDNN and TENSORRT)
-* test_mnistRT: the mnist network hardcoded in using tensorRT apis (TENSORRT only)
-* test_yolo: YOLO detection network (CUDNN and TENSORRT)
-* test_yolo_tiny: smaller version of YOLO (CUDNN and TENSRRT)
-* test_yolo3_berkeley: our yolo3 version trained with BDD100K dateset 
+For specific details on how to export weights see [HERE](./docs/exporting_weights.md).
+
+## Run the demos 
+
+For specific details on how to run:
+- 2D object detection demos, details on FP16, INT8 and batching see [HERE](./docs/demo.md).
+- segmentation demos see [HERE](./docs/README_seg.md).
+- monocular depth estimation see [HERE](./docs/README_depth.md).
+- 2D/3D object detection and tracking demos see [HERE](./docs/README_2d3dtracking.md).
+- mAP demo to evaluate 2D object detectors see [HERE](./docs/mAP_demo.md).
 
-## yolo3 berkeley demo detection
-For the live detection you need to precompile the tensorRT file by luncing the desidered network test, this is the recommended process:
-```
-export TKDNN_MODE=FP16   # set the half floating point optimization
-rm yolo3_berkeley.rt		 # be sure to delete(or move) old tensorRT files
-./test_yolo3_berkeley              # run the yolo test (is slow)
-# with f16 inference the result will be a bit incorrect
-```
-this will genereate a yolo3_berkeley.rt file that can be used for live detection:
-```
-./yolo3_demo                               # launch detection on a demo video
-./yolo3_demo yolo3_berkeley.rt /dev/video0 # launch detection on device 0
-```
 ![demo](https://user-images.githubusercontent.com/11562617/72547657-540e7800-388d-11ea-83c6-49dfea2a0607.gif)
+
+## tkDNN on Windows 10 or Windows 11
+
+For specific details on how to run tkDNN on Windows 10/11 see [HERE](./docs/windows.md).
+
+## Existing tests and supported networks
+
+| Test Name         | Network                                       | Dataset                                                       | N Classes | Input size    | Weights                                                                   |
+| :---------------- | :-------------------------------------------- | :-----------------------------------------------------------: | :-------: | :-----------: | :------------------------------------------------------------------------ |
+| yolo              | YOLO v2<sup>1</sup>                           | [COCO 2014](http://cocodataset.org/)                          | 80        | 608x608       | [weights](https://cloud.hipert.unimore.it/s/nf4PJ3k8bxBETwL/download)                                                                   |
+| yolo_224          | YOLO v2<sup>1</sup>                           | [COCO 2014](http://cocodataset.org/)                          | 80        | 224x224       | weights                                                                   |
+| yolo_berkeley     | YOLO v2<sup>1</sup>                           | [BDD100K  ](https://bair.berkeley.edu/blog/2018/05/30/bdd/)   | 10        | 416x736       | weights                                                                   |
+| yolo_relu         | YOLO v2 (with ReLU, not Leaky)<sup>1</sup>    | [COCO 2014](http://cocodataset.org/)                          | 80        | 416x416       | weights                                                                   |
+| yolo_tiny         | YOLO v2 tiny<sup>1</sup>                      | [COCO 2014](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/m3orfJr8pGrN5mQ/download)                                                                   |
+| yolo_voc          | YOLO v2<sup>1</sup>                           | [VOC      ](http://host.robots.ox.ac.uk/pascal/VOC/)          | 21        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/DJC5Fi2pEjfNDP9/download)                                                                   |
+| yolo3             | YOLO v3<sup>2</sup>                           | [COCO 2014](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/jPXmHyptpLoNdNR/download)     |
+| yolo3_512   | YOLO v3<sup>2</sup>                                 | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/RGecMeGLD4cXEWL/download)     |
+| yolo3_berkeley    | YOLO v3<sup>2</sup>                           | [BDD100K  ](https://bair.berkeley.edu/blog/2018/05/30/bdd/)   | 10        | 320x544       | [weights](https://cloud.hipert.unimore.it/s/o5cHa4AjTKS64oD/download)                                                                   |
+| yolo3_coco4       | YOLO v3<sup>2</sup>                           | [COCO 2014](http://cocodataset.org/)                          | 4         | 416x416       | [weights](https://cloud.hipert.unimore.it/s/o27NDzSAartbyc4/download)                                                                   |
+| yolo3_flir        | YOLO v3<sup>2</sup>                           | [FREE FLIR](https://www.flir.com/oem/adas/adas-dataset-form/) | 3         | 320x544       | [weights](https://cloud.hipert.unimore.it/s/62DECncmF6bMMiH/download)                                                                   |
+| yolo3_tiny        | YOLO v3 tiny<sup>2</sup>                      | [COCO 2014](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/LMcSHtWaLeps8yN/download)     |
+| yolo3_tiny512     | YOLO v3 tiny<sup>2</sup>                      | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/8Zt6bHwHADqP4JC/download)     |
+| dla34             | Deep Leayer Aggreagtion (DLA) 34<sup>3</sup>  | [COCO 2014](http://cocodataset.org/)                          | 80        | 224x224       | weights                                                                   |
+| dla34_cnet        | Centernet (DLA34 backend)<sup>4</sup>         | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/KRZBbCQsKAtQwpZ/download)     |
+| mobilenetv2ssd    | Mobilnet v2 SSD Lite<sup>5</sup>              | [VOC      ](http://host.robots.ox.ac.uk/pascal/VOC/)          | 21        | 300x300       | [weights](https://cloud.hipert.unimore.it/s/x4ZfxBKN23zAJQp/download)     |
+| mobilenetv2ssd512 | Mobilnet v2 SSD Lite<sup>5</sup>              | [COCO 2017](http://cocodataset.org/)                          | 81        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/pdCw2dYyHMJrcEM/download)     |
+| resnet101         | Resnet 101<sup>6</sup>                        | [COCO 2014](http://cocodataset.org/)                          | 80        | 224x224       | weights                                                                   |
+| resnet101_cnet    | Centernet (Resnet101 backend)<sup>4</sup>     | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/5BTjHMWBcJk8g3i/download)     |
+| csresnext50-panet-spp    | Cross Stage Partial Network <sup>7</sup>     | [COCO 2014](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/Kcs4xBozwY4wFx8/download)     |
+| yolo4             | Yolov4 <sup>8</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/d97CFzYqCPCp5Hg/download)     |
+| yolo4_320         | Yolov4 <sup>8</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 320x320       | [weights](https://cloud.hipert.unimore.it/s/d97CFzYqCPCp5Hg/download)     |
+| yolo4_512         | Yolov4 <sup>8</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/d97CFzYqCPCp5Hg/download)     |
+| yolo4_608         | Yolov4 <sup>8</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 608x608       | [weights](https://cloud.hipert.unimore.it/s/d97CFzYqCPCp5Hg/download)     |
+| yolo4_berkeley             | Yolov4 <sup>8</sup>                           | [BDD100K  ](https://bair.berkeley.edu/blog/2018/05/30/bdd/)                          | 10        | 544x320       | [weights](https://cloud.hipert.unimore.it/s/nkWFa5fgb4NTdnB/download)     |
+| yolo4tiny             | Yolov4 tiny <sup>9</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 416x416       | [weights](https://cloud.hipert.unimore.it/s/iRnc4pSqmx78gJs/download)     |
+| yolo4x             | Yolov4x-mish  <sup>9</sup>                          | [COCO 2017](http://cocodataset.org/)                          | 80        | 640x640       | [weights](https://cloud.hipert.unimore.it/s/5MFjtNtgbDGdJEo/download)     |
+| yolo4tiny_512           | Yolov4 tiny <sup>9</sup>                           | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/iRnc4pSqmx78gJs/download)     |
+| yolo4x-cps            | Scaled Yolov4 <sup>10</sup>                          | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/AfzHE4BfTeEm2gH/download)     |
+| shelfnet              | ShelfNet18_realtime<sup>11</sup>                      | [Cityscapes](https://www.cityscapes-dataset.com/)                          | 19        | 1024x1024       | [weights](https://cloud.hipert.unimore.it/s/mEDZMRJaGCFWSJF/download)                                                                   |
+| shelfnet_berkeley              | ShelfNet18_realtime<sup>11</sup>                           | [DeepDrive](https://bdd-data.berkeley.edu/)                          | 20        | 1024x1024       | [weights](https://cloud.hipert.unimore.it/s/m92e7QdD9gYMF7f/download)                                                                   |
+| dla34_cnet3d        | Centernet3D (DLA34 backend)<sup>4</sup>         | [KITTI 2017](http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=3d)                          | 1        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/2MDyWGzQsTKMjmR/download)     |
+| dla34_ctrack        | CenterTrack (DLA34 backend)<sup>12</sup>         | [NuScenes 3D](https://www.nuscenes.org/)                          | 7        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/rjNfgGL9FtAXLHp/download)     |
+| monodepth2        | Monodepth2 <sup>13</sup>         | [KITTI DEPTH](http://www.cvlibs.net/datasets/kitti/raw_data.php)                          | -        | 640x192       | [weights-mono](https://cloud.hipert.unimore.it/s/iYw9QwgP6CsqxLR/download)     |
+| monodepth2        | Monodepth2 <sup>13</sup>         | [KITTI DEPTH](http://www.cvlibs.net/datasets/kitti/raw_data.php)                          | -        | 640x192       | [weights-stereo](https://cloud.hipert.unimore.it/s/XmwbWNXDfqyQ4EL/download)     |
+
+
+## References
+
+1. Redmon, Joseph, and Ali Farhadi. "YOLO9000: better, faster, stronger." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
+2. Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement." arXiv preprint arXiv:1804.02767 (2018).
+3. Yu, Fisher, et al. "Deep layer aggregation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
+4. Zhou, Xingyi, Dequan Wang, and Philipp Krähenbühl. "Objects as points." arXiv preprint arXiv:1904.07850 (2019).
+5. Sandler, Mark, et al. "Mobilenetv2: Inverted residuals and linear bottlenecks." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
+6. He, Kaiming, et al. "Deep residual learning for image recognition." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
+7. Wang, Chien-Yao, et al. "CSPNet: A New Backbone that can Enhance Learning Capability of CNN." arXiv preprint arXiv:1911.11929 (2019).
+8. Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. "YOLOv4: Optimal Speed and Accuracy of Object Detection." arXiv preprint arXiv:2004.10934 (2020).
+9. Bochkovskiy, Alexey, "Yolo v4, v3 and v2 for Windows and Linux" (https://github.com/AlexeyAB/darknet)
+10. Wang, Chien-Yao, Alexey Bochkovskiy, and Hong-Yuan Mark Liao. "Scaled-YOLOv4: Scaling Cross Stage Partial Network." arXiv preprint arXiv:2011.08036 (2020).
+11. Zhuang, Juntang, et al. "ShelfNet for fast semantic segmentation." Proceedings of the IEEE International Conference on Computer Vision Workshops. 2019.
+12. Zhou, Xingyi, Vladlen Koltun, and Philipp Krähenbühl. "Tracking objects as points." European Conference on Computer Vision. Springer, Cham, 2020.
+13. Godard, Clément, et al. "Digging into self-supervised monocular depth estimation." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019.
+
+## Contributors
+The main contibutors, in chronological order, are: 
+- [Francesco Gatti](https://github.com/ceccocats), francesco.gatti@hipert.it
+- [Micaela Verucchi](https://github.com/mive93), micaela.verucchi@unimore.it
+- [Davide Sapienza](https://github.com/sapienzadavide), davide.sapienza@unimore.it
+- [Harshvardhan Chandirasekar](https://github.com/perseusdg), f20180523@goa.bits-pilani.ac.in

cmake/FindCUDNN.cmake

@@ -1,33 +1,66 @@
-# Find the header files
+# find the library
+if(CUDA_FOUND)
+  find_cuda_helper_libs(cudnn)
+  set(CUDNN_LIBRARY ${CUDA_cudnn_LIBRARY} CACHE FILEPATH "location of the cuDNN library")
+  unset(CUDA_cudnn_LIBRARY CACHE)
 
-find_path(CUDNN_INCLUDE_DIR
-        ${CMAKE_SYSROOT}/usr/local/include
-        ${CMAKE_SYSROOT}/usr/include
-        /usr/local/nvidia/tensorrt/include/
-        NO_DEFAULT_PATH
-)
+  find_cuda_helper_libs(nvinfer)
+  set(NVINFER_LIBRARY ${CUDA_nvinfer_LIBRARY} CACHE FILEPATH "location of the nvinfer library")
+  unset(CUDA_nvinfer_LIBRARY CACHE)
+endif()
 
-set(OLD_ROOT ${CMAKE_FIND_ROOT_PATH})
-list(APPEND CMAKE_FIND_ROOT_PATH /)
-list(APPEND CMAKE_FIND_LIBRARY_SUFFIXES .so.7)
-list(APPEND CMAKE_FIND_LIBRARY_SUFFIXES .so.5)
-find_library(CUDNN_LIB
-    NAMES cudnn
-    PATHS
-    /usr/local/driveworks/targets/${CMAKE_SYSTEM_PROCESSOR}-Linux/lib
-    /usr/lib/${CMAKE_SYSTEM_PROCESSOR}-linux-gnu/
+# find the include
+if(CUDNN_LIBRARY)
+  find_path(CUDNN_INCLUDE_DIR
+    cudnn.h
+    PATHS ${CUDA_TOOLKIT_INCLUDE}
+    DOC "location of cudnn.h"
     NO_DEFAULT_PATH
-)
-find_library(CUDNN_NVLIB
-    NAMES "nvinfer"
-    PATHS
-    /usr/local/driveworks/targets/${CMAKE_SYSTEM_PROCESSOR}-Linux/lib
-    /usr/lib/${CMAKE_SYSTEM_PROCESSOR}-linux-gnu/
-    NO_DEFAULT_PATH
-)
-set(CMAKE_FIND_ROOT_PATH ${OLD_ROOT})
+  )
 
-set(CUDNN_LIBRARIES ${CUDNN_LIB} ${CUDNN_NVLIB})
-message("-- Found CUDNN: "  ${CUDNN_LIB})
-message("-- Found NVINFER: "  ${CUDNN_NVLIB})
-set(CUDNN_FOUND true)
+  if(NOT CUDNN_INCLUDE_DIR)
+    find_path(CUDNN_INCLUDE_DIR
+      cudnn.h
+      DOC "location of cudnn.h"
+    )
+  endif()
+
+  message("-- Found CUDNN: " ${CUDNN_LIBRARY})
+  message("-- Found CUDNN include: " ${CUDNN_INCLUDE_DIR})
+endif()
+
+if(NVINFER_LIBRARY)
+  find_path(NVINFER_INCLUDE_DIR
+    NvInfer.h
+    PATHS ${CUDA_TOOLKIT_INCLUDE}
+    DOC "location of NvInfer.h"
+    NO_DEFAULT_PATH
+  )
+
+  if(NOT NVINFER_INCLUDE_DIR)
+    find_path(NVINFER_INCLUDE_DIR
+        NvInfer.h
+        DOC "location of NvInfer.h"
+    )
+  endif()
+
+  message("-- Found NVINFER: " ${NVINFER_LIBRARY})
+  message("-- Found NVINFER include: " ${NVINFER_INCLUDE_DIR})
+endif()
+
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(CUDNN
+  FOUND_VAR CUDNN_FOUND
+  REQUIRED_VARS
+  CUDNN_LIBRARY
+  CUDNN_INCLUDE_DIR
+  VERSION_VAR CUDNN_VERSION
+)
+
+if(CUDNN_FOUND)
+  set(CUDNN_LIBRARIES ${CUDNN_LIBRARY} ${NVINFER_LIBRARY})
+  set(CUDNN_INCLUDE_DIRS ${CUDNN_INCLUDE_DIR} ${NVINFER_INCLUDE_DIR})
+endif()
+
+set(CUDNN_FOUND true)

cmake/getCudaArch.cu

@@ -0,0 +1,20 @@
+#include <stdio.h>
+
+int main(int argc, char **argv){
+    cudaDeviceProp dP;
+    float min_cc = 5.0;
+
+    int rc = cudaGetDeviceProperties(&dP, 0);
+    if(rc != cudaSuccess) {
+        cudaError_t error = cudaGetLastError();
+        printf("CUDA error: %s", cudaGetErrorString(error));
+        return rc; /* Failure */
+    }
+    if((dP.major+(dP.minor/10)) < min_cc) {
+        printf("Min Compute Capability of %2.1f required:  %d.%d found\n Not Building CUDA Code", min_cc, dP.major, dP.minor);
+        return 1; /* Failure */
+    } else {
+        printf("-arch=sm_%d%d", dP.major, dP.minor);
+        return 0; /* Success */
+    }
+}

demo/config.yaml

@@ -0,0 +1,7 @@
+classes     : 80    #number of classes
+map_points  : 101   #number of recall points (0 for all, 101 for COCO, 11 PascalVOC)
+map_levels  : 10    #number of IoU step for the AP
+map_step    : 0.05  #step of IoU
+IoU_thresh  : 0.5   #starting IoU threshold
+conf_thresh : 0.001   #threshold on the condifence of the bbox
+verbose     : false #print on screen information

demo/config_tetra.yaml

@@ -0,0 +1,7 @@
+classes     : 3    #number of classes
+map_points  : 101   #number of recall points (0 for all, 101 for COCO, 11 PascalVOC)
+map_levels  : 10    #number of IoU step for the AP
+map_step    : 0.05  #step of IoU
+IoU_thresh  : 0.5   #starting IoU threshold
+conf_thresh : 0.0   #threshold on the condifence of the bbox
+verbose     : false #print on screen information

demo/demo/data/calib36.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Mon 06 May 2019 11:32:13 PM CEST"
-image_width: 1920
-image_height: 1080
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 1.6158690952190570e+03, 0., 9.4702812371722337e+02, 0.,
-       1.6123979985757153e+03, 5.1995630055718266e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -4.0971199964304100e-01, 1.8755404192050384e-01,
-       -5.3059322427743867e-03, -1.0380603625304912e-03, 0. ]
-avg_reprojection_error: 3.4351035832972515e-01

demo/demo/data/calib_cameras/10218.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:32:53 AM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 4.7264390181579711e+03, 0., 1.5059850098280642e+03, 0.,
-       4.6793092340700096e+03, 6.7300681982359868e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -4.2669569210605879e-01, 6.6337608795749903e-01,
-       -1.3881256269106437e-03, 5.2468063845700682e-03, 0. ]
-avg_reprojection_error: 3.1312290189919406e-01

demo/demo/data/calib_cameras/20932.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:14:44 AM CEST"
-image_width: 1920
-image_height: 1080
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 1.6902498656747011e+03, 0., 9.7959318966703324e+02, 0.,
-       1.7552884617253583e+03, 5.3327953707582492e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -5.4891909767312119e-01, 2.5555919841568631e-01,
-       -4.3831358875660656e-03, -1.3934378903760349e-02, 0. ]
-avg_reprojection_error: 1.1758482932800183e+00

demo/demo/data/calib_cameras/20936.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Mon 06 May 2019 11:32:13 PM CEST"
-image_width: 1920
-image_height: 1080
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 1.6158690952190570e+03, 0., 9.4702812371722337e+02, 0.,
-       1.6123979985757153e+03, 5.1995630055718266e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -4.0971199964304100e-01, 1.8755404192050384e-01,
-       -5.3059322427743867e-03, -1.0380603625304912e-03, 0. ]
-avg_reprojection_error: 3.4351035832972515e-01

demo/demo/data/calib_cameras/20937.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 09:56:50 AM CEST"
-image_width: 1920
-image_height: 1080
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 1.6229477302581809e+03, 0., 1.0277357980566628e+03, 0.,
-       1.6485741394129034e+03, 5.5596919291027621e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.7853584845653426e-01, 7.8553352913896368e-02,
-       -6.5552938633907229e-03, -1.6436824648695104e-02, 0. ]
-avg_reprojection_error: 8.4629096638637347e-01

demo/demo/data/calib_cameras/20939.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Fri 03 May 2019 11:56:13 PM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 5.8796921906556563e+03, 0., 1.3036708932691290e+03, 0.,
-       5.9435402023228071e+03, 8.1110067822514861e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -5.4688862790206871e-01, 5.1913397860290666e-01,
-       -2.1076612628273591e-03, 1.6869796115416984e-02, 0. ]
-avg_reprojection_error: 6.7667474319420251e-01

demo/demo/data/calib_cameras/20940.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Sat 04 May 2019 12:00:38 AM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 4.6903033136815602e+03, 0., 1.6303445000881884e+03, 0.,
-       4.7582671272189546e+03, 4.3596515032334111e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.4366857232996317e-01, 2.2799325522263861e-01,
-       2.0765840315530557e-02, -4.0088654509745098e-03, 0. ]
-avg_reprojection_error: 3.9811872397860709e-01

demo/demo/data/calib_cameras/6310.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:20:53 AM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 2.5005410461483498e+03, 0., 1.5319405824251596e+03, 0.,
-       2.5001544574623872e+03, 7.8267345299919543e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.7379752112038928e-01, 1.6246299444310250e-01,
-       8.0371978716752837e-04, -9.6108499236087584e-04, 0. ]
-avg_reprojection_error: 3.6334262234685299e-01

demo/demo/data/calib_cameras/6311.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:37:47 AM CEST"
-image_width: 960
-image_height: 720
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 5.0439587680799593e+02, 0., 4.8997081391816727e+02, 0.,
-       5.0714582349015507e+02, 3.5481348085748095e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -2.7916140864065331e-01, 6.5465070220501562e-02,
-       -1.9231901334709591e-03, -2.6191562264760264e-03, 0. ]
-avg_reprojection_error: 5.7283635087126605e-01

demo/demo/data/calib_cameras/6312.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:41:12 AM CEST"
-image_width: 960
-image_height: 720
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 5.1663651913150818e+02, 0., 4.7267297458218127e+02, 0.,
-       5.1291090124818436e+02, 3.8505850298928243e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -2.8051872523046845e-01, 6.0895269981008610e-02,
-       -9.7920840355269542e-03, -4.9804820350633240e-04, 0. ]
-avg_reprojection_error: 5.4967308787122626e-01

demo/demo/data/calib_cameras/6313.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:50:02 AM CEST"
-image_width: 960
-image_height: 720
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 4.9724079419911664e+02, 0., 4.9277930193807083e+02, 0.,
-       4.9700744926387819e+02, 3.6581239154403062e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -2.7582961261093608e-01, 6.6908017283259263e-02,
-       -2.1580546593114500e-03, -1.7921711595441153e-03, 0. ]
-avg_reprojection_error: 3.7129088933918375e-01

demo/demo/data/calib_cameras/6314.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:53:27 AM CEST"
-image_width: 960
-image_height: 720
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 4.9372152821507876e+02, 0., 4.7585791077351445e+02, 0.,
-       4.9644139996881893e+02, 3.5961856724726260e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -2.9023109325424973e-01, 8.3150964750672046e-02,
-       -6.1378621304345154e-04, 8.4481910933416999e-04, 0. ]
-avg_reprojection_error: 3.4691001942524069e-01

demo/demo/data/calib_cameras/6315.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Sat 04 May 2019 12:35:58 AM CEST"
-image_width: 3840
-image_height: 2160
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 2.1723071272381276e+03, 0., 1.9718118689531000e+03, 0.,
-       2.2377541672328439e+03, 9.3157209524899565e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.8516162509048857e-01, 1.8961757063227327e-01,
-       1.8297248985443184e-02, -8.9166274086698288e-03, 0. ]
-avg_reprojection_error: 9.9886914863900311e-01

demo/demo/data/calib_cameras/632.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Sun 05 May 2019 08:52:13 PM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 2.9841357325808735e+03, 0., 1.5379802472694901e+03, 0.,
-       2.9784613885271938e+03, 8.9330228722164566e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -5.1096356919758967e-01, 1.4543132746407733e-01,
-       -3.1254001577433334e-02, -1.4769334036191385e-02, 0. ]
-avg_reprojection_error: 9.3544537534095662e-01

demo/demo/data/calib_cameras/634.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Sun 05 May 2019 09:24:33 PM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 2.4796388675592771e+03, 0., 1.5358283835422017e+03, 0.,
-       2.4440198814655632e+03, 8.9911455540136217e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.9478938399374452e-01, 1.6288159087710818e-01,
-       -1.8565610712959927e-02, -7.0112574756757643e-03, 0. ]
-avg_reprojection_error: 4.5805459213906724e-01

demo/demo/data/calib_cameras/637.params

@@ -1,22 +0,0 @@
-%YAML:1.0
----
-calibration_time: "Tue 07 May 2019 10:03:44 AM CEST"
-image_width: 3072
-image_height: 1728
-board_width: 8
-board_height: 6
-square_size: 2.4799999237060547e+01
-flags: 0
-camera_matrix: !!opencv-matrix
-   rows: 3
-   cols: 3
-   dt: d
-   data: [ 2.6027348174982544e+03, 0., 1.4808496083807213e+03, 0.,
-       2.6008830910556521e+03, 6.7577068120137187e+02, 0., 0., 1. ]
-distortion_coefficients: !!opencv-matrix
-   rows: 5
-   cols: 1
-   dt: d
-   data: [ -3.4912899377320661e-01, 1.5704840296202566e-01,
-       6.4926875404798358e-03, 5.7293259996249049e-03, 0. ]
-avg_reprojection_error: 4.0040122960491076e-01

demo/demo/data/pmat_geo/pmat_geo_10218.txt

@@ -1,3 +0,0 @@
--7.327951322519857 -89.11104695972499 9201.012307315486 
--3.2739347567332553 -42.14373356189513 3668.5149400959867 
--0.0009310035308911926 -0.010684335913469737 0.9999999999999999 

demo/demo/data/pmat_geo/pmat_geo_20932.txt

@@ -1,3 +0,0 @@
--5.575118981764042 19.15613757077127 7116.209593619794 
--3.2016502214672586 11.087267568165238 4660.384266971425 
--0.0007553414127200226 0.002676057142671632 1.0 

demo/demo/data/pmat_geo/pmat_geo_20936.txt

@@ -1,3 +0,0 @@
--0.6209787370656474 -57.192772931734694 6553.189887121747 
--0.655304496315561 -28.195100762686465 3423.351597211246 
--0.0001275987755308679 -0.007884524572727637 1.0 

demo/demo/data/pmat_geo/pmat_geo_20937.txt

@@ -1,3 +0,0 @@
-21.441694474392808 72.45500438933801 4652.4523150318 
-9.195810184174585 36.76655098601718 1801.6321216143722 
-0.0027471614789484795 0.009610371445379061 1.0 

demo/demo/data/pmat_geo/pmat_geo_20939.txt

@@ -1,3 +0,0 @@
-23.89713091621312 60.523880117656084 7778.591483347838 
-11.708054761838232 41.819183783017614 1126.9348661897782 
-0.0029807026524197224 0.008783281479831609 0.9999999999999999 

demo/demo/data/pmat_geo/pmat_geo_20940.txt

@@ -1,3 +0,0 @@
--2.986713098145805 -21.641128011999868 6870.903103366085 
--2.2937864581163416 -14.410613254705751 4594.877638262097 
--0.0004752252182089383 -0.003112785793525571 1.0 

demo/demo/data/pmat_geo/pmat_geo_6310.txt

@@ -1,3 +0,0 @@
-0.797119441300711 35.495698816749574 2225.384878217832 
-0.3510443326339413 59.809425152982726 1419.4212456173686 
--1.8197506232485688e-05 0.01820976131616732 1.0 

demo/demo/data/pmat_geo/pmat_geo_6311.txt

@@ -1,3 +0,0 @@
-2.213864892663308 -21.831117643802187 4060.61306610222 
-0.9075230321117119 -7.525265027425343 1268.217122301812 
-0.0005431549026467866 -0.0054930502425022745 1.0 

demo/demo/data/pmat_geo/pmat_geo_6312.txt

@@ -1,3 +0,0 @@
--8.65061527352736 -23.247834477892475 4009.9545358742134 
--2.2093309523396223 -8.398154406654037 1529.4606465341913 
--0.001952580524058866 -0.006089308376531831 1.0 

demo/demo/data/pmat_geo/pmat_geo_6313.txt

@@ -1,3 +0,0 @@
--8.665173353556622 -19.056833294770822 3827.027102449301 
--2.728678410673174 -6.695956964089541 1523.1971944608492 
--0.0020218867007343213 -0.004877696027397352 1.0 

demo/demo/data/pmat_geo/pmat_geo_6314.txt

@@ -1,3 +0,0 @@
--0.5624066154780419 -26.36968028701744 3384.7920236393156 
--0.6590389521668137 -8.709883019993336 1162.7049573736972 
--0.00023421305979212285 -0.006824826866955653 0.9999999999999999 

demo/demo/data/pmat_geo/pmat_geo_6315.txt

@@ -1,3 +0,0 @@
--0.9691844827251948 -16.63913290283819 3991.384826153086 
--0.47672518258283236 -11.365071723693541 1770.7201067235435 
--0.0003215600914247826 -0.005142032883105236 1.0 

demo/demo/data/pmat_geo/pmat_geo_634.txt

@@ -1,3 +0,0 @@
--1.5172087045737013 5.226213011053854 2493.822558690447 
--1.4892268949636678 4.265359557115566 3382.185465939972 
--0.000492856899218649 0.0016917382715537009 1.0 

demo/demo/data/pmat_geo/pmat_geo_636.txt

@@ -1,3 +0,0 @@
-3.671630456405992 102.56885593938215 3570.381264275829 
-1.5985035927893978 46.57491080817888 -1491.4874345505793 
-0.0006446393888135914 0.027894339637393125 1.0 

demo/demo/data/pmat_geo/pmat_geo_637.txt

@@ -1,3 +0,0 @@
-0.44666021794890903 -32.932455985363575 3557.76647014547 
-1.0444994361599629 -37.23760709667856 4404.4517853552425 
-0.00029114742043039743 -0.010721230871024154 1.0 

demo/demo/data/pmundist.txt

@@ -1,3 +0,0 @@
-0.1982518156851575	-13.22038303683631	595.3838470607648	
--0.4806350794284377	-6.567788488139482	598.1732743332492	
--0.00016595895699412456	-0.007499178419813009	1.0

demo/demo/data/proj_matrix_map_b.txt

@@ -1,3 +0,0 @@
--0.4122368700442484 -10.479982981650977 812.1821012303 
--0.6536846365005352 -4.951476039607947 512.0831604767536 
--0.00042630219831388434 -0.006003223898658603 1.0 

demo/demo/demo.cpp

@@ -1,603 +1,156 @@
-#include <time.h>
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+//#include <unistd.h>
+#include <mutex>
 
-#include "utils.h"
+#include "CenternetDetection.h"
+#include "MobilenetDetection.h"
 #include "Yolo3Detection.h"
-#include "message.h"
-#include "visualization.h"
-#include "configuration.h"
-
-#include "tracker.h"
-#include "../masa_protocol/include/send.hpp"
-#include "../masa_protocol/include/serialize.hpp"
-
-// #include <assert.h>
-// #include <unistd.h>
-// #include <mutex>
-// #include <ctime>
-// #include <pthread.h>
-// #include <signal.h>
-// #include <chrono>
-// #include <math.h>
-// #include <typeinfo>
-// #include <iostream>
-
-#define MAX_DETECT_SIZE 100
 
 bool gRun;
-std::chrono::steady_clock::time_point local_clock_start;
-std::mutex mutexgRun;
-std::string obj_class[10]{"person", "car", "truck", "bus", "motor", "bike", "rider", "traffic light", "traffic sign", "train"};
-//mutex for some opencv operations
-std::mutex mutex_cv;
-Show_t updates;
-bool SAVE_RESULT = false;
 
-void sig_handler(int signo)
-{
-    std::cout << "request gateway stop\n";
-    mutexgRun.lock();
+void sig_handler(int signo) {
+    std::cout<<"request gateway stop\n";
     gRun = false;
-    mutexgRun.unlock();
 }
 
-void *readVideoCapture(void *x_void_ptr)
-{
-    std::cout << "readVideoCapture start...\n";
+int main(int argc, char *argv[]) {
 
-    Frame_t *info_f = (Frame_t *)x_void_ptr;
-    mutex_cv.lock();
-    cv::VideoCapture cap(info_f->input, cv::CAP_FFMPEG);
-    mutex_cv.unlock();
-    cv::Mat frame_loc, frame0;
-    int frame_nbr_loc = 0;
-    // bool to_show = false;
-    if (!cap.isOpened())
-    {
-        mutexgRun.lock();
-        gRun = false;
-        mutexgRun.unlock();
-    }
-    else
-        std::cout << "camera started\n";
+    signal(SIGINT, sig_handler);
 
-    // cap.set(cv::CAP_PROP_BUFFERSIZE,3);
-    // std::cout<<"buf size: "<<cap.get(CV_CAP_PROP_BUFFERSIZE)<<std::endl;
-    auto start_t = std::chrono::steady_clock::now();
-    auto step_t = std::chrono::steady_clock::now();
-    auto end_t = std::chrono::steady_clock::now();
-    auto current_timestamp = std::chrono::steady_clock::now();
-
-    // compute fps and find camera's clock
-    double shift, mean_time = 0;
-    std::cout << "Frames per second using video.get(cv::CAP_PROP_FPS) : " << cap.get(cv::CAP_PROP_FPS) << std::endl;
-    std::cout << "readVideoCapture computes frame rate...\n";
-    // //compute frame rate
-    int i = 0;
-    int num_f = 120;
-    // the first 20 frames are null
-    while (i < 21)
-    {
-        cap >> frame_loc;
-        i++;
-    }
-
-    i = 0;
-    start_t = std::chrono::steady_clock::now();
-    while (i < num_f)
-    {
-        step_t = std::chrono::steady_clock::now();
-        cap >> frame_loc;
-        mean_time = mean_time + std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::steady_clock::now() - step_t).count();
-        std::cout << " step " << i << " : " << std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::steady_clock::now() - step_t).count() << " ms" << std::endl;
-        i++;
-    }
-    end_t = std::chrono::steady_clock::now();
-
-    std::cout << "Capturing " << num_f << " frames" << std::endl;
-    std::cout << " Time taken : " << std::chrono::duration_cast<std::chrono::milliseconds>(end_t - start_t).count() << " ms" << std::endl;
-
-    /*
-    mean_time indicates the milliseconds from a frame and the next. (frame rate)
-    local_clock_sync is the camera clock.
-    shift is the difference from local camera clock and local process clock.
-    a frame is allowed if its local timestamp minus its local clock is less then a tollerance, 
-    otherwise it will be considered old.
-    */
-    auto local_clock_sync = std::chrono::steady_clock::now();
-    mean_time = mean_time / num_f;
-    shift = ((double)std::chrono::duration_cast<std::chrono::milliseconds>(local_clock_sync - local_clock_start).count()) / mean_time;
-    shift = (shift - (int)shift) * mean_time;
-    std::cout << ".-------------------------------\n";
-    std::cout << " mean time: " << mean_time << std::endl;
-    std::cout << " shift: " << shift << std::endl;
-    std::cout << " TIMEDIFFERENCE:  " << std::chrono::duration_cast<std::chrono::milliseconds>(local_clock_sync - local_clock_start).count() << std::endl;
-    std::cout << "\n\n\n\n";
-    std::cout << "readVideoCapture start to capture...\n";
-    while (gRun)
-    {
-        // mutex_cv.lock();
-        cap >> frame_loc;
-        // mutex_cv.unlock();
-        current_timestamp = std::chrono::steady_clock::now();
-        shift = std::chrono::duration_cast<std::chrono::milliseconds>(current_timestamp - local_clock_sync).count();
-        std::cout << " RELATIVE TIMESTAMP FRAME : " << shift << " ms" << std::endl;
-        shift = shift / mean_time;
-        shift = (shift - (int)shift) * mean_time;
-        shift = (shift - mean_time / 2 >= 0) ? -(mean_time - shift) : shift;
-        std::cout << "DELAY frame_" << frame_nbr_loc << " : " << shift << " ms" << std::endl;
-        // TODO: here introduce a tollerance to discard old frame
-
-        // std::cout<< "CV_CAP_PROP_POS_MSEC:   "<< cap.get( cv::CAP_PROP_POS_MSEC) <<std::endl;
-        // std::cout<< "CV_CAP_PROP_POS_FRAMES:  "<< cap.get( cv::CAP_PROP_POS_FRAMES) <<std::endl;  // <-- the v4l2 'sequence' field
-        // std::cout<< "cv::CAP_PROP_FPS:  "<< cap.get( cv::CAP_PROP_FPS)<<std::endl;
-        // std::cout << "Format: " << cap.get(CV_CAP_PROP_FORMAT) << "\n";
-        // CAP_PROP_POS_MSEC Current position of the video file in milliseconds or video capture timestamp.
-        std::cout << "id: " << cap.get(cv::CAP_PROP_POS_MSEC) << std::endl;
-        // CAP_PROP_FRAME_COUNT Number of frames in the video file.
-        std::cout << "id: " << cap.get(cv::CAP_PROP_FRAME_COUNT) << std::endl;
-
-        if (!frame_loc.data)
-        {
-            usleep(1000000);
-            mutex_cv.lock();
-            cap.open(info_f->input);
-            printf("cap reinitialize\n");
-            mutex_cv.unlock();
-            continue;
-        }
-
-        end_t = std::chrono::steady_clock::now();
-        std::cout << " VC-TIME 1 : " << std::chrono::duration_cast<std::chrono::milliseconds>(end_t - start_t).count() << " ms" << std::endl;
-        start_t = end_t;
-
-        info_f->sem_vc.lock();
-        info_f->frame = frame_loc.clone();
-        info_f->frame_nbr = frame_nbr_loc;
-        info_f->sem_vc.unlock();
-        // usleep(50000);
-        end_t = std::chrono::steady_clock::now();
-        std::cout << " VC-TIME 2 : " << std::chrono::duration_cast<std::chrono::milliseconds>(end_t - start_t).count() << " ms" << std::endl;
-        start_t = end_t;
-        frame_nbr_loc++;
-    }
-    return (void *)0;
-}
-
-void *computationTask(void *x_void_ptr)
-{
-    Camera_t *camera = (Camera_t *)x_void_ptr;
-    pthread_t visual, originalshow, detectionshow, topviewshow, disparityshow;
-    pthread_t videocap;
-    tk::dnn::Yolo3Detection yolo = *(camera->yolo);
-    //create video capture thread
-    Frame_t info_f;
-    info_f.input = camera->input;
-    if (pthread_create(&videocap, NULL, readVideoCapture, (void *)&info_f))
-    {
-        fprintf(stderr, "Error creating thread\n");
-        return (void *)1;
-    };
-
-    bool to_show = camera->to_show;
-    double adfGeoTransform[6];
-    for (int i = 0; i < 6; i++)
-        adfGeoTransform[i] = camera->adfGeoTransform[i];
-
-    ModFrame_t info_show;
-    if (to_show)
-    {
-        // initialize updates struct
-        updates.update_o = false;
-        updates.update_de = false;
-        updates.update_t = false;
-        updates.update_di = false;
-        if (pthread_create(&visual, NULL, show_updates, (void *)NULL))
-        {
-            fprintf(stderr, "Error creating thread\n");
-            return (void *)1;
-        };
-        if (pthread_create(&originalshow, NULL, originalFrame, (void *)&info_f))
-        {
-            fprintf(stderr, "Error creating thread\n");
-            return (void *)1;
-        };
-        if (pthread_create(&disparityshow, NULL, disparityFrame, (void *)&info_f))
-        {
-            fprintf(stderr, "Error creating thread\n");
-            return (void *)1;
-        };
-        info_show.H = cv::Mat(cv::Size(3, 3), CV_64FC1);
-        if (pthread_create(&detectionshow, NULL, detectionFrame, (void *)&info_show))
-        {
-            fprintf(stderr, "Error creating thread\n");
-            return (void *)1;
-        };
-        if (pthread_create(&topviewshow, NULL, topviewFrame, (void *)&info_show))
-        {
-            fprintf(stderr, "Error creating thread\n");
-            return (void *)1;
-        };
-    }
-    char *pmatrix = camera->pmatrix;
-    /*projection matrix from camera to map*/
-    cv::Mat H(cv::Size(3, 3), CV_64FC1);
-    read_projection_matrix(H, pmatrix);
-    assert(cv::countNonZero(H) > 0);
-    // std::cout<<H<<std::endl;
-    // return (void*)0;
-    /*Camera calibration*/
-    cv::Mat cameraMat, distCoeff;
-    readCameraCalibrationYaml(camera->cameraCalib, cameraMat, distCoeff);
-    std::cout << cameraMat << std::endl;
-    std::cout << distCoeff << std::endl;
-
-    /*GPS information*/
-    std::vector<ObjCoords> coords;
-
-    /*socket*/
-    Communicator Comm(SOCK_DGRAM);
-    Comm.open_client_socket((char *)"127.0.0.1", 8888);
-
-    Message *m = new Message;
-    m->cam_idx = camera->CAM_IDX;
-    m->lights.clear();
-    /*Conversion for tracker, from gps to meters and viceversa*/
-    // mutex_cv.lock();
-    geodetic_converter::GeodeticConverter gc;
-    gc.initialiseReference(44.655540, 10.934315, 0);
-    // mutex_cv.unlock();
-    double east, north, up;
-    // double lat, lon, alt;
-    /*Mask info*/
-    cv::Mat mask = cv::imread(camera->maskfile, cv::IMREAD_GRAYSCALE);
-    cv::Mat maskOrient = cv::imread(camera->maskFileOrient);
-    // cv::Mat maskOrient = cv::imread(camera->maskFileOrient, 0);
-
-    /*for(int i=0; i< mask.cols; i++)
-    {
-        for(int j=0; j< mask.rows; j++)
-        {
-            std::cout<<maskOrient.at<cv::Vec3b>(i,j)    <<std::endl;
-        }
-    }
+    // get config file path and read it
+    #ifdef __linux__ 
+        std::string config_file = "../demo/demoConfig.yaml";
+    #elif _WIN32
+        std::string config_file = "..\\..\\..\\demo\\demoConfig.yaml";
+    #endif
+    if(argc > 1)
+        config_file = argv[1]; 
     
-   
-    return 0;*/
-    /*tracker infos*/
-    std::vector<Tracker> trackers;
-    std::vector<Data> cur_frame;
-    int initial_age = -5;
-    int age_threshold = -8;
-    int n_states = 5;
-    float dt = 0.03;
+    YAML::Node conf =  YAMLloadConf(config_file);
+    if(!conf)
+        FatalError("Problem with config file");
 
-    int frame_nbr = 0;
+    // read settings from config file
+    std::string net = YAMLgetConf<std::string>(conf, "net", "yolo4tiny_fp32.rt");
+    if(!fileExist(net.c_str()))
+        FatalError("The given network does not exist. Create the rt first.");
 
-    //save video
-    /*cv::VideoWriter outputVideo;
-    cv::Size S = cv::Size((int)cap.get(cv::CAP_PROP_FRAME_WIDTH), //Acquire input size
-                          (int)cap.get(cv::CAP_PROP_FRAME_HEIGHT));
-    outputVideo.open("test.avi", static_cast<int>(cap.get(cv::CAP_PROP_FOURCC)), cap.get(cv::CAP_PROP_FPS), S, true);*/
+    #ifdef __linux__
+        std::string input = YAMLgetConf<std::string>(conf, "input", "../demo/yolo_test.mp4");
+    #elif _WIN32
+        std::string input = YAMLgetConf<std::string>(conf, "win_input", "..\\..\\..\\demo\\yolo_test.mp4");
+    #endif
+        if(!fileExist(input.c_str()))
+        FatalError("The given input video does not exist.");
 
-    cv::Mat map1, map2;
-    auto start_t = std::chrono::steady_clock::now();
-    auto step_t = std::chrono::steady_clock::now();
-    auto end_t = std::chrono::steady_clock::now();
-    // auto step_t_segmentation = std::chrono::steady_clock::now();
-    // auto end_t_segmentation = std::chrono::steady_clock::now();
+    char ntype          = YAMLgetConf<char>(conf, "ntype", 'y');
+    int n_classes       = YAMLgetConf<int>(conf, "n_classes", 80);
+    int n_batch         = YAMLgetConf<int>(conf, "n_batch", 1);
+    if(n_batch < 1 || n_batch > 64)
+        FatalError("Batch dim not supported");
+    float conf_thresh   = YAMLgetConf<float>(conf, "conf_thresh", 0.3);
+    bool show           = YAMLgetConf<bool>(conf, "show", true);
+    bool save           = YAMLgetConf<bool>(conf, "save", false);
 
-    //TODO: move in a thread
-    // // information for the disparity map
-    // std::vector <cv::Rect> pre_rois;
-    // cv::Mat pre_frame;
-    cv::Mat orig_frame;
-    // cv::Mat canny, pre_canny, canny_RGB, pre_canny_RGB;
-    // cv::Mat canny_img;
+    std::cout   <<"Net settings - net: "<< net
+                <<", ntype: "<< ntype
+                <<", n_classes: "<< n_classes
+                <<", n_batch: "<< n_batch
+                <<", conf_thresh: "<< conf_thresh<<"\n"; 
+    std::cout   <<"Demo settings - input: "<< input
+                <<", show: "<< show
+                <<", save: "<< save<<"\n\n"; 
+    
+    // create detection network
+    tk::dnn::Yolo3Detection yolo;
+    tk::dnn::CenternetDetection cnet;
+    tk::dnn::MobilenetDetection mbnet;
 
-    // box variable
-    tk::dnn::box b;
-    int x0, h, y1; //w, x1, y0;
-    int objClass;
-    std::string det_class;
-    ;
-    // float prob;
-    cv::Scalar intensity;
+    tk::dnn::DetectionNN *detNN;  
 
+    switch(ntype)
+    {
+        case 'y':
+            detNN = &yolo;
+            break;
+        case 'c':
+            detNN = &cnet;
+            break;
+        case 'm':
+            detNN = &mbnet;
+            n_classes++;
+            break;
+        default:
+        FatalError("Network type not allowed (3rd parameter)\n");
+    }
 
-    // cv::VideoWriter resultVideo;
-    // if(SAVE_RESULT) {
-    //     int w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
-    //     int h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);
-    //     resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
-    // }
+    detNN->init(net,n_classes,n_batch,conf_thresh);
+
+    // open video stream
+    cv::VideoCapture cap(input);
+    if(!cap.isOpened())
+        gRun = false; 
+    else
+        std::cout<<"camera started\n";
+
+    cv::VideoWriter resultVideo;
+    if(save) {
+        int w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
+        int h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);
+        resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
+    }
+
+    if(show)
+        cv::namedWindow("detection", cv::WINDOW_NORMAL);
 
     cv::Mat frame;
-    cv::Mat frame_crop;
-    cv::Mat dnn_input;
-    bool first_iteration = true;
+    std::vector<cv::Mat> batch_frame;
+    std::vector<cv::Mat> batch_dnn_input;
 
-    while (gRun)
-    {
-        TIMER_START
-        start_t = std::chrono::steady_clock::now();
-        step_t = start_t;
+    // start detection loop
+    gRun = true;
+    while(gRun) {
+        batch_dnn_input.clear();
+        batch_frame.clear();
+        
+        for(int bi=0; bi< n_batch; ++bi){
+            cap >> frame; 
+            if(!frame.data) 
+                break;
+            
+            batch_frame.push_back(frame);
 
-        info_f.sem_vc.lock();
-        frame = info_f.frame.clone();
-        if (info_f.frame_nbr - frame_nbr > 1)
-            std::cout << "more than one - f_n (diff " << info_f.frame_nbr - frame_nbr << ")\n";
-        frame_nbr = info_f.frame_nbr;
-        info_f.sem_vc.unlock();
-        std::cout << "f_n: " << frame_nbr << std::endl;
-        // if (!frame.data)
-        if (frame_nbr == 0)
-        {
-            usleep(1000000);
-            printf("no frame received\n");
-            continue;
-        }
-        orig_frame = frame.clone();
-        // mutex_cv.lock();
-        if (first_iteration)
-            cv::initUndistortRectifyMap(cameraMat, distCoeff, cv::Mat(), cameraMat, frame.size(), CV_16SC2, map1, map2);
-        cv::Mat temp = frame.clone();
-        cv::remap(temp, frame, map1, map2, 1);
-        //undistort(temp, frame, cameraMat, distCoeff);
-        // mutex_cv.unlock();
+            // this will be resized to the net format
+            batch_dnn_input.push_back(frame.clone());
+        } 
+        if(!frame.data) 
+            break;
+    
+        //inference
+        detNN->update(batch_dnn_input, n_batch);
+        detNN->draw(batch_frame);
 
-        // this will be resized to the net format
-        dnn_input = frame.clone();
-        // TODO: async infer
-        yolo.update(dnn_input);
-        int num_detected = yolo.detected.size();
-        if (num_detected > MAX_DETECT_SIZE)
-            num_detected = MAX_DETECT_SIZE;
-
-        coords.clear();
-        end_t = std::chrono::steady_clock::now();
-        std::cout << " TIME 1 : " << std::chrono::duration_cast<std::chrono::milliseconds>(end_t - step_t).count() << " ms" << std::endl;
-        step_t = end_t;
-        // draw dets
-        std::cout << "camera: " << camera->CAM_IDX << " - num detected: " << num_detected << std::endl;
-
-        //TODO: move in a thread
-        // //preprocessing frame
-        // step_t_segmentation = std::chrono::steady_clock::now();
-        // // src_gray
-        // canny_img = img_laplacian(orig_frame,0);
-        // cv::Canny(canny_img, canny, 100, 100*2 );
-        // // sprintf(buf_frame_crop_name,"../demo/demo/data/img_disparity/%d_%d_canny.jpg",frame_nbr, 999);
-        // // cv::imwrite(buf_frame_crop_name, canny);
-        // end_t_segmentation = std::chrono::steady_clock::now();
-        // std::cout << " - TIME END pre canny : "<<std::chrono::duration_cast<std::chrono::milliseconds>(end_t_segmentation - step_t_segmentation).count() << " ms"<<std::endl;
-        // step_t_segmentation = end_t_segmentation;
-        // // std::cout<<"o: "<<orig_frame.cols<<" - "<<orig_frame.rows<<std::endl;
-        // // std::cout<<"canny: "<<canny.cols<<" - "<<canny.rows<<std::endl;
-        // // std::cout<<"pre: "<<pre_canny.cols<<" - "<<pre_canny.rows<<std::endl;
-        // if(!first_iteration)
-        // {
-        //     // backtorgb = cv::cvtColor(pre_canny,cv::COLOR_GRAY2RGB)
-        //     cv::cvtColor(pre_canny, pre_canny_RGB, cv::COLOR_GRAY2RGB);
-        //     cv::cvtColor(canny, canny_RGB, cv::COLOR_GRAY2RGB);
-        //     disparity_frame = frame_disparity(pre_canny_RGB, canny_RGB, frame_nbr, 999, 0);
-        //     std::cout<<"size: "<<disparity_frame.rows<<" - "<<disparity_frame.cols<<std::endl;
-        //     if (disparity_frame.rows == 0 || disparity_frame.cols == 0)
-        //         return -1;
-        //     if (disparity_frame.empty())
-        //     { // only fools don't check...
-        //         std::cout << "image not loaded !" << std::endl;
-        //         return -1;
-        //     }
-        //     end_t_segmentation = std::chrono::steady_clock::now();
-        //     std::cout << " TIME canny : frame_disparity : "<<std::chrono::duration_cast<std::chrono::milliseconds>(end_t_segmentation - step_t_segmentation).count() << " ms"<<std::endl;
-        //     step_t_segmentation = end_t_segmentation;
-
-        //     // //--------------------------------
-        //     // //frame box disparity on the original image
-        //     // step_t_segmentation = std::chrono::steady_clock::now();
-        //     // frame_box_disparity(pre_frame, frame, pre_rois, frame_nbr);
-        //     // // reset pre_rois for the new roi of the current frame
-        //     // // pre_rois.erase(pre_rois.begin(), pre_rois.end());
-        //     // end_t_segmentation = std::chrono::steady_clock::now();
-        //     // std::cout << " TIME Frame disparity : "<<std::chrono::duration_cast<std::chrono::milliseconds>(end_t_segmentation - step_t_segmentation).count() << " ms"<<std::endl;
-        //     // step_t_segmentation = end_t_segmentation;
-
-        //     // //frame box disparity on the preprocessed image
-        //     // cv::cvtColor(pre_canny, pre_canny_RGB, cv::COLOR_GRAY2RGB);
-        //     // cv::cvtColor(canny, canny_RGB, cv::COLOR_GRAY2RGB);
-        //     // frame_box_disparity(pre_canny_RGB, canny_RGB, pre_rois, frame_nbr);
-        //     // // reset pre_rois for the new roi of the current frame
-        //     // pre_rois.erase(pre_rois.begin(), pre_rois.end());
-        //     // end_t_segmentation = std::chrono::steady_clock::now();
-        //     // std::cout << " TIME Canny Frame disparity : "<<std::chrono::duration_cast<std::chrono::milliseconds>(end_t_segmentation - step_t_segmentation).count() << " ms"<<std::endl;
-        //     // step_t_segmentation = end_t_segmentation;
-        //     // //---------------------------------
-        // }
-
-        // compute some metrics on the whole frame
-        // segmentation(pre_frame, frame, frame_nbr, 0, 0);
-
-        for (int i = 0; i < num_detected; i++)
-        {
-            b = yolo.detected[i];
-            x0 = b.x;
-            // w = b.w;
-            // x1 = b.x + w;
-            // y0 = b.y;
-            h = b.h;
-            y1 = b.y + h;
-            objClass = b.cl;
-            det_class = obj_class[b.cl];
-            // prob = b.prob;
-
-            intensity = mask.at<uchar>(cv::Point(int(x0 + b.w / 2), y1));
-
-            if (intensity[0])
-            {
-
-                if (objClass < 6)
-                {
-
-                    // find the rectangular on the frame (sub-figure)
-                    // roi.x = (x0 > 0)? x0 : 0;
-                    // roi.y = (y0 > 0)? y0 : 0;
-                    // // std::cout<<"x "<<roi.x<<" - y "<<roi.y<<std::endl;
-                    // roi.width = (roi.x+w >= frame.cols)? frame.cols-1-roi.x : w;
-                    // roi.height = (roi.y+h >= frame.rows)? frame.rows-1-roi.y : h;
-                    // std::cout<<"w "<<roi.width<<" - h "<<roi.height<<std::endl;
-                    // std::cout<<"wf "<<frame.cols<<" - hf "<<frame.rows<<std::endl;
-                    // std::cout<<"---"<<std::endl;
-                    // std::cout<<"x "<<roi.x<<" to "<<roi.width+roi.x<<" wf "<<frame.cols<<std::endl;
-                    // std::cout<<"y "<<roi.y<<" to "<<roi.height+roi.y<<" hf "<<frame.rows<<std::endl;
-                    //update pre_roi for the next frame
-                    // pre_rois.push_back(roi);
-
-                    // segmentation(frame(roi), frame(roi), frame_nbr, i, 1);
-
-                    /////
-                    convert_coords(coords, x0 + b.w / 2, y1, objClass, H, adfGeoTransform);
-
-                    // //std::cout<<objClass<<" ("<<prob<<"): "<<x0<<" "<<y0<<" "<<x1<<" "<<y1<<"\n";
-                    // cv::rectangle(frame, cv::Point(x0, y0), cv::Point(x1, y1), camera->yolo.colors[objClass], 2);
-                    // // draw label
-                    // int baseline = 0;
-                    // float fontScale = 0.5;
-                    // int thickness = 2;
-                    // cv::Size textSize = getTextSize(det_class, cv::FONT_HERSHEY_SIMPLEX, fontScale, thickness, &baseline);
-                    // cv::rectangle(frame, cv::Point(x0, y0), cv::Point((x0 + textSize.width - 2), (y0 - textSize.height - 2)), camera->yolo.colors[b.cl], -1);
-                    // cv::putText(frame, det_class, cv::Point(x0, (y0 - (baseline / 2))), cv::FONT_HERSHEY_SIMPLEX, fontScale, cv::Scalar(255, 255, 255), thickness);
-                }
+        if(show){
+            for(int bi=0; bi< n_batch; ++bi){
+                cv::imshow("detection", batch_frame[bi]);
+                cv::waitKey(1);
             }
         }
-
-        end_t = std::chrono::steady_clock::now();
-        std::cout << " TIME 2 : " << std::chrono::duration_cast<std::chrono::milliseconds>(end_t - step_t).count() << " ms" << std::endl;
-        step_t = end_t;
-        //convert from latitude and longitude to meters for ekf
-        cur_frame.clear();
-        for (size_t i = 0; i < coords.size(); i++)
-        {
-            gc.geodetic2Enu(coords[i].lat_, coords[i].long_, 0, &east, &north, &up);
-            cur_frame.push_back(Data(east, north, frame_nbr, coords[i].class_));
-        }
-        if (first_iteration)
-        {
-            // if there aren't detections and it is the first iteration, we can't initialize the tracker, so continue
-            if (cur_frame.empty())
-                continue;
-            for (auto f : cur_frame)
-                trackers.push_back(Tracker(f, initial_age, dt, n_states));
-        }
-        else
-        {
-            Track(cur_frame, dt, n_states, initial_age, age_threshold, trackers);
-        }
-        std::cout << "There are " << trackers.size() << " trackers" << std::endl;
-        //prepare message with tracker info
-        if (trackers.size() != 0)
-        {
-            // mutex_cv.lock();
-            addRoadUserfromTracker(trackers, m, gc, maskOrient, adfGeoTransform, H);
-            // mutex_cv.unlock();
-            //prepare the message with detection info
-            //prepare_message(m, coords, CAM_IDX);
-            //send message
-            if (!m->objects.empty())
-                Comm.send_message(m);
-        }
-
-        if (to_show)
-        {
-            //populate the ModFrame_t
-            info_show.sem.lock();
-            info_show.original_frame = frame.clone();
-            // std::vector<Tracker> trackers;
-            info_show.trackers = trackers;
-            // geodetic_converter::GeodeticConverter gc;
-            info_show.gc = gc;
-            for (int i = 0; i < 6; i++)
-                info_show.adfGeoTransform[i] = adfGeoTransform[i];
-            // cv::Mat H;
-            info_show.H = H.clone();
-            info_show.yolo = yolo;
-            // std::copy(camera->yolo.begin(), camera->yolo.end(), info_show.yolo.begin());
-            info_show.mask = mask.clone();
-            info_show.sem.unlock();
-        }
-
-        // update pre_frame for the disparity map
-        // pre_frame = orig_frame.clone();
-        // pre_canny = canny.clone();
-        if (first_iteration)
-            first_iteration = false;
-
-        frame_nbr++;
-        std::cout << camera->CAM_IDX << " camera thread: ";
-        TIMER_STOP
+        if(n_batch == 1 && save)
+            resultVideo << frame;
     }
-    return (void *)0;
+
+    std::cout<<"detection end\n";   
+    
+    double mean = 0; 
+    std::cout<<COL_GREENB<<"\n\nTime stats:\n";
+    std::cout<<"Min: "<<*std::min_element(detNN->stats.begin(), detNN->stats.end())<<" ms\n";
+    std::cout<<"Max: "<<*std::max_element(detNN->stats.begin(), detNN->stats.end())<<" ms\n";    
+    for(int i=0; i<detNN->stats.size(); i++) mean += detNN->stats[i]; mean /= detNN->stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\t"<<1000/(mean)<<" FPS\n"<<COL_END;   
+
+    return 0;
 }
 
-int main(int argc, char *argv[])
-{
-
-    std::cout << "detection\n";
-    signal(SIGINT, sig_handler);
-    srand(time(NULL));
-
-    Parameters_t par;
-
-    if(!read_parameters(argc, argv, &par))
-        return -1;
-    
-    tk::dnn::Yolo3Detection yolo[par.n_cameras];
-    for(int i=0; i<par.n_cameras; i++)
-    {
-        yolo[i].init(par.net);
-        yolo[i].thresh = 0.25;    
-        
-        // if(SAVE_RESULT)
-        //     resultVideo << frame;
-    }
-    // tk::dnn::Yolo3Detection yolo;
-    // yolo.init(net);
-    // yolo.thresh = 0.25;    
-    
-    gRun = true;
-    // start the local clock. It is used to check the incoming frames (by different cameras) 
-    local_clock_start = std::chrono::steady_clock::now();
-
-    /*GPS information*/
-    double *adfGeoTransform = (double *)malloc(6 * sizeof(double));
-    readTiff(par.tiffile, adfGeoTransform);
-    // Camera_t cameras[par.n_cameras]; 
-    for(int i=0; i<par.n_cameras; i++)
-    {
-        for(int j = 0; j < 6; j++ )
-            par.cameras[i].adfGeoTransform[j] = adfGeoTransform[j];
-        par.cameras[i].yolo = &yolo[i];
-        // par.cameras[i].yolo.init(par.net);
-        // par.cameras[i].yolo.thresh = 0.25;
-        // cameras[i].yolo = yolo[i];
-        // cameras[i].yolo = yolo;
-        
-    }
-    pthread_t camera_task[par.n_cameras];
-    for(int i=0; i<par.n_cameras; i++)
-    {
-        std::cout<<"creating thread\n";
-        if(pthread_create(&camera_task[i], NULL, computationTask, (void*)&(par.cameras[i])))
-        {
-            fprintf(stderr, "error creating thread\n");
-            return 1;
-        }
-    }
-    for(int i=0; i<par.n_cameras; i++)
-    {
-        pthread_join(camera_task[i], NULL);
-    }
-    std::cout <<" free adfGeoT \n";
-    free(adfGeoTransform);
-    std::cout << "detection end\n";
-    return 0;
-}

demo/demo/demo3D.cpp

@@ -0,0 +1,159 @@
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+//#include <unistd.h>
+#include <mutex>
+
+#include "demo_utils.h"
+#include "CenternetDetection3D.h"
+
+bool gRun;
+bool SAVE_RESULT = false;
+
+void sig_handler(int signo) {
+    std::cout<<"request gateway stop\n";
+    gRun = false;
+}
+
+int main(int argc, char *argv[]) {
+
+    std::cout<<"detection\n";
+    signal(SIGINT, sig_handler);
+
+
+    std::string net = "dla34_cnet3d_fp32.rt";
+    if(argc > 1)
+        net = argv[1]; 
+    #ifdef __linux__ 
+        std::string input = "../demo/yolo_test.mp4";
+    #elif _WIN32
+        std::string input = "..\\..\\..\\demo\\yolo_test.mp4";
+    #endif
+
+    if(argc > 2)
+        input = argv[2]; 
+    std::string calib_params = "";
+    if(argc > 3)
+        calib_params = argv[3]; 
+    char ntype = 'c';
+    if(argc > 4)
+        ntype = argv[4][0]; 
+    int n_classes = 3;
+    if(argc > 5)
+        n_classes = atoi(argv[5]); 
+    int n_batch = 1;
+    if(argc > 6)
+        n_batch = atoi(argv[6]); 
+        
+    bool show = true;
+    if(argc > 7)
+        show = atoi(argv[7]); 
+    float conf_thresh=0.3;
+    if(argc > 8)
+        conf_thresh = atof(argv[8]);     
+
+    if(n_batch < 1 || n_batch > 64)
+        FatalError("Batch dim not supported");
+
+    if(!show)
+	    SAVE_RESULT = true;
+
+    tk::dnn::CenternetDetection3D cnet;
+
+    tk::dnn::DetectionNN3D *detNN;  
+
+    switch(ntype)
+    {
+        case 'c':
+            detNN = &cnet;
+            break;
+        default:
+        FatalError("Network type not allowed (3rd parameter)\n");
+    }
+    std::vector<cv::Mat> calibs;
+    if(!calib_params.empty() && calib_params!="NULL") {
+        std::cout<<"calib_params: "<<calib_params<<std::endl;
+        cv::Mat calib;
+        // the calibration matrix must be a 3x3 matrix
+        readCalibrationMatrix(calib_params, calib);
+        for(int bi=0; bi< n_batch; ++bi)
+            calibs.push_back(calib);
+    }
+    detNN->init(net, n_classes, n_batch, conf_thresh, calibs);
+
+    gRun = true;
+
+    cv::VideoCapture cap(input);
+    if(!cap.isOpened())
+        gRun = false; 
+    else
+        std::cout<<"camera started\n";
+
+    cv::VideoWriter resultVideo;
+    if(SAVE_RESULT) {
+        int w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
+        int h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);
+        resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
+    }
+    cv::Mat frame;
+    if(show)
+	    cv::namedWindow("detection", cv::WINDOW_NORMAL);
+    
+    std::vector<cv::Mat> batch_frame;
+    std::vector<cv::Mat> batch_dnn_input;
+
+    while(gRun) {
+        batch_dnn_input.clear();
+        batch_frame.clear();
+        
+        for(int bi=0; bi< n_batch; ++bi){
+            cap >> frame; 
+            if(!frame.data) 
+                break;
+            batch_frame.push_back(frame);
+
+            // this will be resized to the net format
+            batch_dnn_input.push_back(frame.clone());
+        } 
+        if(!frame.data) 
+            break;  
+ 
+        //inference
+        detNN->update(batch_dnn_input, n_batch, false, nullptr, false);
+        detNN->draw(batch_frame);
+
+        if(show){
+            for(int bi=0; bi< n_batch; ++bi){
+                cv::imshow("detection", batch_frame[bi]);
+                cv::waitKey(1);
+            }
+        }
+        if(n_batch == 1 && SAVE_RESULT)
+            resultVideo << frame;
+    }
+
+    std::cout<<"detection end\n";   
+    double mean = 0; 
+    
+    std::cout<<COL_GREENB<<"\n\nTime preprocessing stats:\n";
+    std::cout<<"Min: "<<*std::min_element(detNN->pre_stats.begin(), detNN->pre_stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(detNN->pre_stats.begin(), detNN->pre_stats.end())<<" ms\n";    
+    for(int i=0; i<detNN->pre_stats.size(); i++) mean += detNN->pre_stats[i]; mean /= detNN->pre_stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    mean=0;
+    std::cout<<COL_GREENB<<"\n\nTime stats:\n";
+    std::cout<<"Min: "<<*std::min_element(detNN->stats.begin(), detNN->stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(detNN->stats.begin(), detNN->stats.end())<<" ms\n";    
+    for(int i=0; i<detNN->stats.size(); i++) mean += detNN->stats[i]; mean /= detNN->stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    mean=0;
+    std::cout<<COL_GREENB<<"\n\nTime postprocessing stats:\n";
+    std::cout<<"Min: "<<*std::min_element(detNN->post_stats.begin(), detNN->post_stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(detNN->post_stats.begin(), detNN->post_stats.end())<<" ms\n";    
+    for(int i=0; i<detNN->post_stats.size(); i++) mean += detNN->post_stats[i]; mean /= detNN->post_stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    
+    
+    return 0;
+}
+

demo/demo/demoDepth.cpp

@@ -0,0 +1,106 @@
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+//#include <unistd.h>
+#include <mutex>
+
+#include "tkDNN/DepthNN.h"
+
+bool gRun;
+
+void sig_handler(int signo) {
+    std::cout<<"request gateway stop\n";
+    gRun = false;
+}
+
+int main(int argc, char *argv[]) {
+
+    signal(SIGINT, sig_handler);
+
+    std::string net = "monodepth2_fp32.rt";
+    if(argc > 1)
+        net = argv[1]; 
+    #ifdef __linux__ 
+        std::string input = "../demo/yolo_test.mp4";
+    #elif _WIN32
+        std::string input = "..\\..\\..\\demo\\yolo_test.mp4";
+    #endif
+    if(argc > 2)
+        input = argv[2]; 
+    bool show = true;
+    if(argc > 3)
+        show = atoi(argv[3]); 
+    bool save = true;
+    if(argc > 4)
+        save = atoi(argv[4]); 
+
+    std::cout   <<"Net settings - net: "<< net
+                <<"\n"; 
+    std::cout   <<"Demo settings - input: "<< input
+                <<", show: "<< show
+                <<", save: "<< save<<"\n\n"; 
+    
+    tk::dnn::DepthNN depthNN;  
+
+    // create depth network 
+    int n_batch = 1;
+    depthNN.init(net, n_batch);
+
+    // open video stream
+    cv::VideoCapture cap(input);
+    if(!cap.isOpened())
+        gRun = false; 
+    else
+        std::cout<<"camera started\n";
+
+    cv::VideoWriter resultVideo;
+    if(save) {
+        int w = depthNN.output_w;
+        int h = depthNN.output_h;
+        resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
+    }
+
+    if(show)
+        cv::namedWindow("depth", cv::WINDOW_NORMAL);
+
+    cv::Mat frame;
+    std::vector<cv::Mat> batch_frame;
+    std::vector<cv::Mat> batch_dnn_input;
+
+    // start detection loop
+    gRun = true;
+    while(gRun) {
+        batch_dnn_input.clear();
+        batch_frame.clear();
+
+        //read frame
+        cap >> frame; 
+        if(!frame.data) 
+            break;
+        batch_frame.push_back(frame);
+        batch_dnn_input.push_back(frame.clone());
+    
+        //inference
+        depthNN.update(batch_dnn_input, 1);
+        if(show){
+            cv::imshow("depth", depthNN.depthMats[0]);
+            cv::waitKey(1);
+
+        }
+
+        if(save)
+            resultVideo << depthNN.depthMats[0];
+    }
+
+    std::cout<<"detection end\n";   
+    
+    double mean = 0; 
+    std::cout<<COL_GREENB<<"\n\nTime stats depth:\n";
+    std::cout<<"Min: "<<*std::min_element(depthNN.stats.begin(), depthNN.stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(depthNN.stats.begin(), depthNN.stats.end())<<" ms\n";    
+    for(int i=0; i<depthNN.stats.size(); i++) mean += depthNN.stats[i]; mean /= depthNN.stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\t"<<1000/(mean)<<" FPS\n";   
+
+    return 0;
+}
+

demo/demo/demoTracker.cpp

@@ -0,0 +1,160 @@
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+//#include <unistd.h>
+#include <mutex>
+
+#include "demo_utils.h"
+#include "CenterTrack.h"
+
+bool gRun;
+bool SAVE_RESULT = false;
+
+void sig_handler(int signo) {
+    std::cout<<"request gateway stop\n";
+    gRun = false;
+}
+
+int main(int argc, char *argv[]) {
+
+    std::cout<<"detection\n";
+    signal(SIGINT, sig_handler);
+
+
+    std::string net = "dla34_cnet3d_track_fp32.rt";
+    if(argc > 1)
+        net = argv[1]; 
+    #ifdef __linux__ 
+        std::string input = "../demo/yolo_test.mp4";
+    #elif _WIN32
+        std::string input = "..\\..\\..\\demo\\yolo_test.mp4";
+    #endif
+
+    if(argc > 2)
+        input = argv[2]; 
+    std::string calib_params = "";
+    if(argc > 3)
+        calib_params = argv[3];
+    char ntype = 'c';
+    if(argc > 4)
+        ntype = argv[4][0]; 
+    int n_classes = 3;
+    if(argc > 5)
+        n_classes = atoi(argv[5]); 
+    int n_batch = 1;
+    if(argc > 6)
+        n_batch = atoi(argv[6]); 
+    bool show = true;
+    if(argc > 7)
+        show = atoi(argv[7]); 
+    float conf_thresh=0.3;
+    if(argc > 8)
+        conf_thresh = atof(argv[8]);   
+    bool t3d = true;
+    if(argc > 9)
+        t3d = atoi(argv[9]);
+    if(n_batch < 1 || n_batch > 64)
+        FatalError("Batch dim not supported");
+
+    if(!show)
+	    SAVE_RESULT = true;
+
+    tk::dnn::CenterTrack ctrack;
+
+    tk::dnn::TrackingNN *trackNN;  
+
+    switch(ntype)
+    {
+        case 'c':
+            trackNN = &ctrack;
+            break;
+        default:
+        FatalError("Network type not allowed (3rd parameter)\n");
+    }
+    std::vector<cv::Mat> calibs;
+    if(!calib_params.empty() && calib_params!="NULL") {
+        std::cout<<"calib_params: "<<calib_params<<std::endl;
+        cv::Mat calib;
+        // the calibration matrix must be a 3x3 matrix
+        readCalibrationMatrix(calib_params, calib);
+        for(int bi=0; bi< n_batch; ++bi)
+            calibs.push_back(calib);
+    }
+    trackNN->init(net, n_classes, n_batch, conf_thresh, t3d, calibs);
+
+    gRun = true;
+
+    cv::VideoCapture cap(input);
+    if(!cap.isOpened())
+        gRun = false; 
+    else
+        std::cout<<"camera started\n";
+
+    cv::VideoWriter resultVideo;
+    if(SAVE_RESULT) {
+        int w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
+        int h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);
+        resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
+    }
+    cv::Mat frame;
+    if(show)
+	    cv::namedWindow("detection", cv::WINDOW_NORMAL);
+    
+    std::vector<cv::Mat> batch_frame;
+    std::vector<cv::Mat> batch_dnn_input;
+
+    while(gRun) {
+        batch_dnn_input.clear();
+        batch_frame.clear();
+        
+        for(int bi=0; bi< n_batch; ++bi){
+            cap >> frame; 
+            if(!frame.data) 
+                break;
+            batch_frame.push_back(frame);
+
+            // this will be resized to the net format
+            batch_dnn_input.push_back(frame.clone());
+        } 
+        if(!frame.data) 
+            break;  
+ 
+        //inference
+        trackNN->update(batch_dnn_input, n_batch, false, nullptr, false);
+        trackNN->draw(batch_frame);
+
+        if(show){
+            for(int bi=0; bi< n_batch; ++bi){
+                cv::imshow("detection", batch_frame[bi]);
+                cv::waitKey(1);
+            }
+        }
+        if(n_batch == 1 && SAVE_RESULT)
+            resultVideo << frame;
+    }
+
+    std::cout<<"detection end\n";   
+    double mean = 0; 
+    
+    std::cout<<COL_GREENB<<"\n\nTime preprocessing stats:\n";
+    std::cout<<"Min: "<<*std::min_element(trackNN->pre_stats.begin(), trackNN->pre_stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(trackNN->pre_stats.begin(), trackNN->pre_stats.end())<<" ms\n";    
+    for(int i=0; i<trackNN->pre_stats.size(); i++) mean += trackNN->pre_stats[i]; mean /= trackNN->pre_stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    mean=0;
+    std::cout<<COL_GREENB<<"\n\nTime stats:\n";
+    std::cout<<"Min: "<<*std::min_element(trackNN->stats.begin(), trackNN->stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(trackNN->stats.begin(), trackNN->stats.end())<<" ms\n";    
+    for(int i=0; i<trackNN->stats.size(); i++) mean += trackNN->stats[i]; mean /= trackNN->stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    mean=0;
+    std::cout<<COL_GREENB<<"\n\nTime postprocessing stats:\n";
+    std::cout<<"Min: "<<*std::min_element(trackNN->post_stats.begin(), trackNN->post_stats.end())<<" ms\n";    
+    std::cout<<"Max: "<<*std::max_element(trackNN->post_stats.begin(), trackNN->post_stats.end())<<" ms\n";    
+    for(int i=0; i<trackNN->post_stats.size(); i++) mean += trackNN->post_stats[i]; mean /= trackNN->post_stats.size();
+    std::cout<<"Avg: "<<mean<<" ms\n"<<COL_END;   
+    
+    
+    return 0;
+}
+

demo/demo/map.cpp

@@ -0,0 +1,264 @@
+
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+#ifdef __linux__
+#include <unistd.h>
+#endif
+
+#include <mutex>
+#include "utils.h"
+
+#include <opencv2/core/core.hpp>
+#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/videoio.hpp>
+#include <opencv2/imgproc/imgproc.hpp>
+
+#include "Yolo3Detection.h"
+#include "CenternetDetection.h"
+#include "MobilenetDetection.h"
+
+#include "evaluation.h"
+
+#include <map>
+
+void convertFilename(std::string &filename,const std::string l_folder, const std::string i_folder, const std::string l_ext,const std::string i_ext)
+{
+    filename.replace(filename.find(l_folder),l_folder.length(),i_folder);
+    filename.replace(filename.find(l_ext),l_ext.length(),i_ext);
+}
+
+int main(int argc, char *argv[]) 
+{
+    char ntype = 'y';
+    const char *config_filename = "../demo/config.yaml";
+    const char * net = "yolo4tiny_fp32.rt";
+    const char * labels_path = "../demo/COCO_val2017/all_labels.txt";
+    int n_batches = 1;
+    float confidence_thresh = 0.3;
+    bool show = false;
+    bool write_dets = false;
+    bool write_res_on_file = true;
+    bool write_coco_json = false;
+    int n_images = 5000;
+
+    bool verbose;
+    int classes, map_points, map_levels;
+    float map_step, IoU_thresh, conf_thresh;
+    double vm_total = 0, rss_total = 0;
+    double vm, rss;
+
+    //read args
+    if(argc > 1)
+        net = argv[1]; 
+    if(argc > 2)
+        ntype = argv[2][0];
+    if(argc > 3)
+        labels_path = argv[3]; 
+    if(argc > 4)
+        config_filename = argv[4]; 
+    if(argc > 5)
+        n_batches = atoi(argv[5]); 
+    if(argc > 6)
+        confidence_thresh = atof(argv[6]); 
+
+    std::cout<<"conf t: "<<confidence_thresh<<std::endl;
+
+    //check if files needed exist
+    if(!fileExist(config_filename))
+        FatalError("Wrong config file path.");
+    if(!fileExist(net))
+        FatalError("Wrong net file path.");
+    if(!fileExist(labels_path))
+        FatalError("Wrong labels file path.");
+
+    //read mAP parameters
+    tk::dnn::readmAPParams( config_filename, classes,  map_points, map_levels, map_step,
+                IoU_thresh, conf_thresh, verbose);
+
+    //extract network name from rt path
+    std::string net_name;
+    removePathAndExtension(net, net_name);
+    std::cout<<"Network: "<<net_name<<std::endl;
+
+    //open files (if needed)
+    std::ofstream times, memory, coco_json;
+
+    if(write_coco_json){
+        coco_json.open(net_name+"_COCO_res.json");
+        coco_json << "[\n";
+    }
+
+    if(write_res_on_file){
+        times.open("times_"+net_name+"_"+ std::to_string(n_batches)+"_"+std::to_string(confidence_thresh)+".csv");
+        memory.open("memory.csv", std::ios_base::app);
+        memory<<net_name+"_"+ std::to_string(n_batches)+"_"+std::to_string(confidence_thresh)<<";";
+    }
+
+    // instantiate detector
+    tk::dnn::Yolo3Detection yolo;
+    tk::dnn::CenternetDetection cnet;
+    tk::dnn::MobilenetDetection mbnet;
+    tk::dnn::DetectionNN *detNN;  
+    int n_classes = classes;   
+    switch(ntype){
+        case 'y':
+            detNN = &yolo;
+            break;
+        case 'c':
+            detNN = &cnet;
+            break;
+        case 'm':
+            detNN = &mbnet;
+            n_classes++;
+            break;
+        default:
+            FatalError("Network type not allowed (3rd parameter)\n");
+    }
+    detNN->init(net,n_classes, 1, conf_thresh);
+
+    //read images 
+    std::ifstream all_labels(labels_path);
+    std::string l_filename;
+    std::vector<tk::dnn::Frame> images;
+    std::vector<tk::dnn::box> detected_bbox;
+
+    std::cout<<"Reading groundtruth and generating detections"<<std::endl;
+
+    if(show)
+        cv::namedWindow("detection", cv::WINDOW_NORMAL);
+
+    bool file_ok = false;
+
+    int images_done;
+    for (images_done=0 ; images_done < n_images ;) {
+        
+
+        int cur_batches = 0;
+        std::vector<cv::Mat> batch_frames;
+        std::vector<cv::Mat> batch_dnn_input;
+            
+        std::vector<tk::dnn::Frame> cur_frames;
+        for(;cur_batches<n_batches && images_done < n_images;cur_batches++, ++images_done){
+
+            std::getline(all_labels, l_filename);
+            file_ok = all_labels ? true : false ;
+            if (!file_ok)
+                break;
+
+            tk::dnn::Frame f;
+            f.lFilename = l_filename;
+            f.iFilename = l_filename;
+            convertFilename(f.iFilename, "labels", "images", ".txt", ".jpg");
+
+            // read frame
+            if(!fileExist(f.iFilename.c_str()))
+                FatalError("Wrong image file path.");
+
+            cv::Mat frame = cv::imread(f.iFilename.c_str(), cv::IMREAD_COLOR);
+            batch_frames.push_back(frame);
+            f.height = frame.rows;
+            f.width = frame.cols;
+
+            if(!frame.data) 
+                break;
+            batch_dnn_input.push_back(frame.clone());
+
+            // read and save groundtruth labels
+            if(fileExist(f.lFilename.c_str()))
+            {
+                std::ifstream labels(f.lFilename);
+                for(std::string line; std::getline(labels, line); ){
+                    std::istringstream in(line); 
+                    tk::dnn::BoundingBox b;
+                    in >> b.cl >> b.x >> b.y >> b.w >> b.h;  
+                    b.prob = 1;
+                    b.truthFlag = 1;
+                    f.gt.push_back(b);
+
+                    if(show)// draw rectangle for groundtruth
+                        cv::rectangle(batch_frames[cur_batches], cv::Point((b.x-b.w/2)*f.width, (b.y-b.h/2)*f.height), cv::Point((b.x+b.w/2)*f.width,(b.y+b.h/2)*f.height), cv::Scalar(0, 255, 0), 2);             
+                }
+            } 
+
+            cur_frames.push_back(f);
+        }
+        if (!file_ok)
+            break;
+
+        //inference 
+        detNN->update(batch_dnn_input,cur_batches,write_res_on_file, &times, write_coco_json);
+        detNN->draw(batch_frames);
+
+        for(int j=0;j<cur_frames.size(); ++j){
+            if(write_coco_json)
+            printJsonCOCOFormat(&coco_json, cur_frames[j].iFilename.c_str(), detNN->batchDetected[j], classes,  cur_frames[j].width, cur_frames[j].height);        
+
+            std::ofstream myfile;
+            if(write_dets)
+                myfile.open ("det/"+cur_frames[j].lFilename.substr(cur_frames[j].lFilename.find("labels/") + 7));
+
+            // save detections labels
+            for(auto d:detNN->batchDetected[j]){
+                //convert detected bb in the same format as label
+                //<x_center>/<image_width> <y_center>/<image_width> <width>/<image_width> <height>/<image_width>
+                tk::dnn::BoundingBox b;
+                b.x = (d.x + d.w/2) / cur_frames[j].width;
+                b.y = (d.y + d.h/2) / cur_frames[j].height;
+                b.w = d.w / cur_frames[j].width;
+                b.h = d.h / cur_frames[j].height;
+                b.prob = d.prob;
+                b.cl = d.cl;
+                cur_frames[j].det.push_back(b);
+
+                if(write_dets)
+                    myfile << d.cl << " "<< d.prob << " "<< b.x << " "<< b.y << " "<< b.w << " "<< b.h <<"\n";
+
+                if(show)// draw rectangle for detection
+                    cv::rectangle(batch_frames[j], cv::Point(d.x, d.y), cv::Point(d.x + d.w, d.y + d.h), cv::Scalar(0, 0, 255), 2);             
+                }
+
+            if(write_dets)
+                myfile.close();
+      
+            images.push_back(cur_frames[j]);
+        
+            if(show){
+                cv::imshow("detection", batch_frames[j]);
+                cv::waitKey(0);
+            }
+            
+        }
+        std::cout <<COL_ORANGEB<< "Images done:\t" << images_done<< "\tcur batch:\t"<<cur_batches<< "\n"<<COL_END;
+        
+        getMemUsage(vm, rss);
+        vm_total += vm;
+        rss_total += rss;
+
+        
+    }
+
+    if(write_coco_json){
+        coco_json.seekp (coco_json.tellp() - std::streampos(2));
+        coco_json << "\n]\n";
+        coco_json.close();
+    }
+
+    std::cout << "Avg VM[MB]: " << vm_total/images_done/1024.0 << ";Avg RSS[MB]: " << rss_total/images_done/1024.0 << std::endl;
+
+    //compute mAP
+    double AP = tk::dnn::computeMapNIoULevels(images,classes,IoU_thresh,confidence_thresh, map_points, map_step, map_levels, verbose, write_res_on_file, net_name+"_"+ std::to_string(n_batches)+"_"+std::to_string(confidence_thresh));
+    std::cout<<"mAP "<<IoU_thresh<<":"<<IoU_thresh+map_step*(map_levels-1)<<" = "<<AP<<std::endl;
+
+    //compute average precision, recall and f1score
+    tk::dnn::computeTPFPFN(images,classes,IoU_thresh,confidence_thresh, verbose, write_res_on_file, net_name +"_"+ std::to_string(n_batches)+"_"+std::to_string(confidence_thresh));
+
+    if(write_res_on_file){
+        memory<<vm_total/images_done/1024.0<<";"<<rss_total/images_done/1024.0<<"\n";
+        times.close();
+        memory.close();
+    }
+
+    return 0;
+}
+

demo/demo/seg_demo.cpp

@@ -0,0 +1,150 @@
+#include <iostream>
+#include <signal.h>
+#include <stdlib.h>     /* srand, rand */
+#ifdef __linux__
+#include <unistd.h>
+#endif 
+#include <mutex>
+
+#include "SegmentationNN.h"
+
+bool gRun;
+bool SAVE_RESULT = true;
+
+void sig_handler(int signo) {
+    std::cout<<"request gateway stop\n";
+    gRun = false;
+}
+
+void writePred(const std::string& images_names, const std::string& gt_folder, const std::string& out_folder, tk::dnn::SegmentationNN& segNN, int& width, int& height, bool show=false){
+    std::ifstream all_gt(images_names);
+    std::string filename;
+    cv::Mat frame;
+    for (; std::getline(all_gt, filename); ) {
+        std::cout<<filename<<std::endl;
+        frame = cv::imread(gt_folder + filename);
+        height = frame.rows;
+        width = frame.cols;
+        segNN.updateOriginal(frame, false);
+        if(show)
+            segNN.draw();
+        cv::imwrite(out_folder + filename, segNN.segmented[0]);
+    }
+}
+
+int main(int argc, char *argv[]) {
+
+    std::cout<<"detection\n";
+    signal(SIGINT, sig_handler);
+
+
+    std::string net = "shelfnet_fp32.rt";
+    if(argc > 1)
+        net = argv[1]; 
+    std::string input = "../demo/yolo_test.mp4";
+    if(argc > 2)
+        input = argv[2]; 
+    int n_batch = 1;
+    if(argc > 3)
+        n_batch = atoi(argv[3]); 
+    int n_classes = 19;
+    if(argc > 4)
+        n_classes = atoi(argv[4]); 
+    bool resize = false;
+    if(argc > 5)
+        resize = atoi(argv[5]); 
+    int baseline_resize = 1024;
+    if(argc > 6)
+        baseline_resize = atoi(argv[6]); 
+    bool show = true;
+    if(argc > 7)
+        show = atoi(argv[7]); 
+    bool write_pred = false;
+    if(argc > 8)
+        write_pred = atoi(argv[8]); 
+
+    if(resize && (baseline_resize < 0 || baseline_resize > 5000)) 
+        FatalError("Problem with baseline resize")
+    if(n_batch < 1 || n_batch > 64)
+        FatalError("Batch dim not supported");
+
+    //net initialization 
+    tk::dnn::SegmentationNN segNN;
+    segNN.init(net, n_classes, n_batch);
+
+    int height = 0, width = 0;
+    int basewidth=baseline_resize,  hsize; 
+    
+    if(write_pred){
+        std::string gt_folder = "../demo/CityScapes_val/images/";
+        std::string images_names = "../demo/CityScapes_val/all_images.txt";
+        std::string out_folder = "seg/";
+
+        writePred(images_names, gt_folder, out_folder, segNN, width, height, show);
+    }
+    else{
+        if(!show)
+            SAVE_RESULT = true;
+
+        gRun = true;
+
+        cv::VideoCapture cap(input);
+        if(!cap.isOpened())
+            gRun = false; 
+        else
+            std::cout<<"camera started\n";
+
+        cv::VideoWriter resultVideo;
+        if(SAVE_RESULT) {
+            int w,h;
+            if(resize){
+                w = basewidth;
+                h = int((float(cap.get(cv::CAP_PROP_FRAME_HEIGHT))*float(basewidth/float(cap.get(cv::CAP_PROP_FRAME_WIDTH)))));
+            }
+            else{
+                w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
+                h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);
+            }
+            resultVideo.open("result.mp4", cv::VideoWriter::fourcc('M','P','4','V'), 30, cv::Size(w, h));
+        }
+
+        cv::Mat frame;
+        while(gRun) {
+            cap >> frame; 
+            if(!frame.data) 
+                break;
+
+            if(resize){
+                hsize  = int((float(frame.rows)*float(basewidth/float(frame.cols))));
+                cv::resize(frame, frame, cv::Size(basewidth, hsize));
+            }
+
+            height = frame.rows;
+            width = frame.cols;
+
+            //inference
+            segNN.updateOriginal(frame, true);
+            if(show)
+                segNN.draw();
+
+            if(SAVE_RESULT)
+                resultVideo << segNN.segmented[0];
+        }
+    }
+
+    std::cout<<"segmentation end\n";   
+    double mean = 0, mean_pre = 0, mean_post = 0;
+    
+    std::cout<<COL_GREENB<<"\n\nTime stats for size ["<<width<<","<<height<<"] :\n";
+    
+    for(int i=0; i<segNN.stats.size(); i++) mean += segNN.stats[i]; mean /= segNN.stats.size();
+    for(int i=0; i<segNN.stats_pre.size(); i++) mean_pre += segNN.stats_pre[i]; mean_pre /= segNN.stats_pre.size();
+    for(int i=0; i<segNN.stats_post.size(); i++) mean_post += segNN.stats_post[i]; mean_post /= segNN.stats_post.size();
+    std::cout<<"Avg pre:\t"<<mean_pre<<" ms\t"<<1000/(mean_pre)<<" FPS\n";   
+    std::cout<<"Avg inf:\t"<<mean<<" ms\t"<<1000/(mean)<<" FPS\n";   
+    std::cout<<"Avg post:\t"<<mean_post<<" ms\t"<<1000/(mean_post)<<" FPS\n\n";   
+    std::cout<<"Avg tot:\t"<<(mean_pre + mean_post + mean) <<" ms\t"<<1000/((mean_pre + mean_post + mean))<<" FPS\n"<<COL_END;   
+    
+    return 0;
+}
+

demo/demoConfig.yaml

@@ -0,0 +1,14 @@
+# video input
+input     : "../demo/yolo_test.mp4"
+win_input : "..\\..\\..\\demo\\yolo_test.mp4"
+
+# network config
+net         : "yolo4_berkeley_fp32.rt"
+ntype       : 'y'
+n_classes   : 80
+n_batch     : 1
+conf_thresh : 0.3
+
+# demo config
+show  : true
+save  : false

docker/Dockerfile

@@ -0,0 +1,7 @@
+FROM ceccocats/tkdnn:latest
+LABEL maintainer "Francesco Gatti"
+
+RUN cd && git clone https://github.com/ceccocats/tkDNN.git && cd tkDNN && mkdir build && cd build \ 
+    && cmake .. && make -j12
+
+

docker/Dockerfile.base

@@ -0,0 +1,140 @@
+FROM nvidia/cudagl:11.3.1-devel-ubuntu20.04
+
+LABEL maintainer "TKDNN AUTHORS"
+LABEL Description="tkDNN+cudagl"
+LABEL com.tkdnn.nvidia.version="11.3.1"
+
+ENV DEBIAN_FRONTEND noninteractive
+ENV CC gcc
+ENV CXX g++
+
+RUN apt-get update && apt-get install -y \
+libblkid-dev && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+RUN apt-get update && apt-get install -y \
+libcudnn8-dev=8.2.1.32-1+cuda11.3 \
+libcudnn8=8.2.1.32-1+cuda11.3 \
+libnvinfer-dev=8.0.3-1+cuda11.3 \
+libnvinfer8=8.0.3-1+cuda11.3 && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+libblkid-dev \
+locales \
+lsb-release \
+mesa-utils \
+git \
+nano \
+terminator \
+wget \
+curl \
+libssl-dev \
+htop \
+dbus-x11 \
+libqt5opengl5-dev \
+libgtk-3-dev \
+libvtk7-dev \
+libv4l-dev \
+tar \
+libgoogle-glog-dev \
+libgflags-dev \
+gfortran-9 \
+libtbb-dev \
+libgstreamer1.0-dev \
+libgstreamer-plugins-base1.0-dev \
+libdc1394-22-dev \
+libavresample-dev \
+libatlas-cpp-0.6-dev \
+python3-dev \
+gdb \
+python3-pip \
+unzip libtbb-dev && \
+apt-get clean && rm -rf /var/lib/apt/lists/*
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+software-properties-common && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+RUN apt-add-repository universe
+RUN apt-get update && apt-get install -y python3-pip  python3 openssh-server ssh pyqt5-dev sip-dev && apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN pip3 install --upgrade pip
+RUN pip3 install --upgrade virtualenv
+RUN pip3 install --upgrade paramiko
+RUN pip3 install --ignore-installed --upgrade numpy protobuf
+
+
+RUN cd ~ && mkdir build
+RUN cd ~/build && wget https://github.com/Kitware/CMake/releases/download/v3.21.4/cmake-3.21.4.tar.gz  && \
+tar -xvf cmake-3.21.4.tar.gz && cd cmake-3.21.4 && ./configure --prefix=/usr/local --qt-gui --parallel=12 && \
+make -j8 && make install
+
+RUN apt-get update && apt-get install -y automake autoconf pkg-config libevent-dev libncurses5-dev bison && \
+apt-get clean && rm -rf /var/lib/apt/lists/
+
+RUN git clone https://github.com/tmux/tmux.git && \
+cd tmux && git checkout tags/3.2 && ls -la && sh autogen.sh && ./configure && make -j8 && make install
+
+RUN apt-get update && apt-get install -y zsh && apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN wget https://github.com/robbyrussell/oh-my-zsh/raw/master/tools/install.sh -O - | zsh || true
+RUN chsh -s /usr/bin/zsh root
+RUN git clone https://github.com/sindresorhus/pure /root/.oh-my-zsh/custom/pure
+RUN ln -s /root/.oh-my-zsh/custom/pure/pure.zsh-theme /root/.oh-my-zsh/custom/
+RUN ln -s /root/.oh-my-zsh/custom/pure/async.zsh /root/.oh-my-zsh/custom/
+RUN sed -i -e 's/robbyrussell/refined/g' /root/.zshrc
+RUN sed -i '/plugins=(/c\plugins=(git git-flow adb pyenv tmux)' /root/.zshrc
+
+RUN mkdir -p /root/.config/terminator/
+COPY assets/terminator_config /root/.config/terminator/config
+
+RUN echo "/usr/local/nvidia/lib" >> /etc/ld.so.conf.d/nvidia.conf && \
+    echo "/usr/local/nvidia/lib64" >> /etc/ld.so.conf.d/nvidia.conf && \
+    echo "/usr/local/cuda/lib64" >> /etc/ld.so.conf.d/nvidia.conf
+
+
+ENV PATH /usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
+ENV LD_LIBRARY_PATH /usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/lib:/usr/lib/x86_64-linux-gnu:/usr/local/lib:${LD_LIBRARY_PATH}
+ENV NVIDIA_VISIBLE_DEVICES all
+ENV NVIDIA_DRIVER_CAPABILITIES compute,utility,graphics
+
+
+
+RUN cd ~/build && wget https://github.com/opencv/opencv/archive/4.5.4.tar.gz && tar -xf 4.5.4.tar.gz && rm 4.5.4.tar.gz
+RUN cd ~/build && wget https://github.com/opencv/opencv_contrib/archive/4.5.4.tar.gz && tar -xf 4.5.4.tar.gz && rm 4.5.4.tar.gz
+RUN cd ~/build && \
+    cd opencv-4.5.4 && mkdir build && cd build && \
+    cmake -D CMAKE_BUILD_TYPE=RELEASE \
+    -D CMAKE_INSTALL_PREFIX=/usr/local \
+    -D INSTALL_PYTHON_EXAMPLES=OFF \
+    -D INSTALL_C_EXAMPLES=OFF \
+    -D OPENCV_EXTRA_MODULES_PATH='~/build/opencv_contrib-4.5.4/modules' \
+    -D BUILD_EXAMPLES=OFF \
+    -D BUILD_TESTS=OFF \
+    -D BUILD_PERF_TESTS=OFF \
+    -D BUILD_DOCS=OFF \
+    -D WITH_CUDA=ON \
+    -D WITH_OPENGL=ON \
+    -D WITH_NVCUVID=ON \
+    -D CUDA_ARCH_BIN=7.2 \
+    -D CUDA_ARCH_PTX=7.2 \
+    -D ENABLE_FAST_MATH=ON \
+    -D CUDA_FAST_MATH=ON \
+    -D WITH_CUBLAS=ON \
+    -D WITH_CUDNN=ON \
+    -D WITH_OPENMP=ON \
+    -D WITH_NONFREE=ON \
+    -D WITH_LIBV4L=ON \
+    -D WITH_GSTREAMER=ON \
+    -D WITH_GSTREAMER_0_10=OFF \
+    -D WITH_TBB=ON \
+    ../ && make -j12 && make install && ldconfig
+
+RUN cd ~ && rm -rf build
+
+RUN cd ~ && mkdir Development && cd Development && \
+git clone https://github.com/ceccocats/tkDNN.git && cd tkDNN && \
+mkdir build && cd build && \
+cmake -DCMAKE_BUILD_TYPE=Release .. && \
+make -j6
+
+RUN apt-get clean && rm -rf /var/lib/apt/lists/*
+COPY assets/entrypoint_setup.sh /
+ENTRYPOINT ["/entrypoint_setup.sh"]
+CMD ["terminator"]

docker/README.md

@@ -0,0 +1,18 @@
+# Use the prebuilt image
+```
+# build image
+docker build -t tkdnn:build -f Dockerfile . 
+```
+
+# Build Base Docker image
+```
+# make nvidia docker working
+# follow this guide: https://github.com/NVIDIA/nvidia-docker
+
+# build image
+docker build -t ceccocats/tkdnn:latest -f Dockerfile.base .
+
+# run image
+./docker_launch.sh
+```
+

docker/assets/entrypoint_setup.sh

@@ -0,0 +1,123 @@
+#! /bin/bash
+
+CMD=
+
+# Functions
+# TOOD: Check if we can use: getent passwd $USER to extract all variables
+# TODO: Check for valid inputs, cause now it will go through even with bad inputs
+check_envs () {
+    DOCKER_CUSTOM_USER_OK=true;
+    if [ -z ${DOCKER_USER_NAME+x} ]; then 
+        DOCKER_CUSTOM_USER_OK=false;
+        return;
+    fi
+    
+    if [ -z ${DOCKER_USER_ID+x} ]; then 
+        DOCKER_CUSTOM_USER_OK=false;
+        return;
+    else
+        if ! [ -z "${DOCKER_USER_ID##[0-9]*}" ]; then 
+            echo -e "\033[1;33mWarning: User-ID should be a number. Falling back to defaults.\033[0m"
+            DOCKER_CUSTOM_USER_OK=false;
+            return;
+        fi
+    fi
+    
+    if [ -z ${DOCKER_USER_GROUP_NAME+x} ]; then 
+        DOCKER_CUSTOM_USER_OK=false;
+        return;
+    fi
+
+    if [ -z ${DOCKER_USER_GROUP_ID+x} ]; then 
+        DOCKER_CUSTOM_USER_OK=false;
+        return;
+    else
+        if ! [ -z "${DOCKER_USER_GROUP_ID##[0-9]*}" ]; then 
+            echo -e "\033[1;33mWarning: Group-ID should be a number. Falling back to defaults.\033[0m"
+            DOCKER_CUSTOM_USER_OK=false;
+            return;
+        fi
+    fi
+}
+
+setup_env_user () {
+    USER=$1
+    USER_ID=$2
+    GROUP=$3
+    GROUP_ID=$4
+
+    ## Create user
+    useradd -m $USER
+
+    ## Copy zsh/sh configs
+    cp /root/.profile /home/$USER/
+    cp /root/.bashrc /home/$USER/
+    cp /root/.zshrc /home/$USER/
+    ## Copy terminator configs
+    mkdir -p /home/$USER/.config/terminator
+    cp /root/.config/terminator/config /home/$USER/.config/terminator/config
+    cp /root/.config/terminator/background.png /home/$USER/.config/terminator/background.png
+    cp -rf /root/.oh-my-zsh /home/$USER/
+    cp -rf /root/tkDNN /home/$USER/
+    rm -rf /home/$USER/.oh-my-zsh/custom/pure.zsh-theme /home/$USER/.oh-my-zsh/custom/async.zsh
+    ln -s /home/$USER/.oh-my-zsh/custom/pure/pure.zsh-theme /home/$USER/.oh-my-zsh/custom/
+    ln -s /home/$USER/.oh-my-zsh/custom/pure/async.zsh /home/$USER/.oh-my-zsh/custom/
+    sed -i -e 's@ZSH=\"/root@ZSH=\"/home/$USER@g' /home/$USER/.zshrc
+    # Copy SSH keys & fix owner
+    if [ -d "/root/.ssh" ]; then
+        cp -rf /root/.ssh /home/$USER/
+        chown -R $USER:$GROUP /home/$USER/.ssh
+    fi
+    
+    ## Fix owner
+    chown $USER:$GROUP /home/$USER
+    chown -R $USER:$GROUP /home/$USER/.config
+    chown $USER:$GROUP /home/$USER/.profile
+    chown $USER:$GROUP /home/$USER/.bashrc
+    chown $USER:$GROUP /home/$USER/.zshrc
+    chown -R $USER:$GROUP /home/$USER/.oh-my-zsh
+    chown -R $USER:$GROUP /home/$USER/tkDNN
+
+    ## This a trick to keep the evnironmental variables of root which is important!
+    echo "if ! [ \"$DOCKER_USER_NAME\" = \"$(id -un)\" ]; then" >> /root/.bashrc
+    echo "    cd /home/$DOCKER_USER_NAME" >> /root/.bashrc
+    echo "    su $DOCKER_USER_NAME" >> /root/.bashrc
+    echo "fi" >> /root/.bashrc
+
+    echo "if ! [ \"$DOCKER_USER_NAME\" = \"$(id -un)\" ]; then" >> /root/.zshrc
+    echo "    cd /home/$DOCKER_USER_NAME" >> /root/.zshrc
+    echo "    su $DOCKER_USER_NAME" >> /root/.zshrc
+    echo "fi" >> /root/.zshrc
+
+    ## Setup Password-file
+    PASSWDCONTENTS=$(grep -v "^${USER}:" /etc/passwd)
+    GROUPCONTENTS=$(grep -v -e "^${GROUP}:" -e "^docker:" /etc/group)
+
+    (echo "${PASSWDCONTENTS}" && echo "${USER}:x:$USER_ID:$GROUP_ID::/home/$USER:/bin/bash") > /etc/passwd
+    (echo "${GROUPCONTENTS}" && echo "${GROUP}:x:${GROUP_ID}:") > /etc/group
+    (if test -f /etc/sudoers ; then echo "${USER}  ALL=(ALL)   NOPASSWD: ALL" >> /etc/sudoers ; fi)
+}
+
+
+# ---Main---
+
+# Create new user
+## Check Inputs
+check_envs
+
+## Determine user & Setup Environment
+if [ $DOCKER_CUSTOM_USER_OK == true ]; then
+    echo "  -->DOCKER_USER Input is set to '$DOCKER_USER_NAME:$DOCKER_USER_ID:$DOCKER_USER_GROUP_NAME:$DOCKER_USER_GROUP_ID'";
+    echo -e "\033[0;32mSetting up environment for user=$DOCKER_USER_NAME\033[0m"
+    setup_env_user $DOCKER_USER_NAME $DOCKER_USER_ID $DOCKER_USER_GROUP_NAME $DOCKER_USER_GROUP_ID
+else
+    echo "  -->DOCKER_USER* variables not set. Using 'root'.";
+    echo -e "\033[0;32mSetting up environment for user=root\033[0m"
+    DOCKER_USER_NAME="root"
+fi
+
+# Change shell to zsh
+chsh -s /usr/bin/zsh $DOCKER_USER_NAME
+
+# Run CMD from Docker
+"$@"

docker/assets/terminator_config

@@ -0,0 +1,18 @@
+[global_config]
+  title_transmit_bg_color = "#2e3436"
+[keybindings]
+[layouts]
+  [[default]]
+    [[[child1]]]
+      parent = window0
+      type = Terminal
+    [[[window0]]]
+      parent = ""
+      type = Window
+[plugins]
+[profiles]
+  [[default]]
+    background_color = "#282828"
+    cursor_color = "#aaaaaa"
+    foreground_color = "#f3f3f3"
+    palette = "#000000:#aa0000:#00aa00:#c4a000:#3465a4:#75507b:#06989a:#d3d7cf:#88807c:#f15d22:#73c48f:#ffce51:#48b9c7:#ad7fa8:#34e2e2:#eeeeec"

docker/docker_launch.sh

@@ -0,0 +1,9 @@
+xhost local:root
+docker run --rm -it --runtime=nvidia --privileged --net=host --cap-add sys_ptrace -d --ipc=host \
+-v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY \
+-v $HOME/.Xauthority:/home/$(id -un)/.Xauthority -e XAUTHORITY=/home/$(id -un)/.Xauthority \
+-e DOCKER_USER_NAME=$(id -un) \
+-e DOCKER_USER_ID=$(id -u) \
+-e DOCKER_USER_GROUP_NAME=$(id -gn) \
+-e DOCKER_USER_GROUP_ID=$(id -g) \
+-v $HOME/.ssh:/home/$(id -un)/.ssh  ceccocats/tkdnn

docs/README_2d3dtracking.md

@@ -0,0 +1,92 @@
+# 2D/3D Object Detection and Tracking
+
+Currently tkDNN supports only CenterTrack as 3DOD & 2D/3D Tracker network.
+
+## 3D Object Detection
+
+To run the 3D object detection demo follow these steps (example with CenterNet based on DLA34):
+```
+rm dla34_cnet3d_fp32.rt        # be sure to delete(or move) old tensorRT files
+./test_dla34_cnet3d            # run the yolo test (is slow)
+./demo3D dla34_cnet3d_fp32.rt ../demo/yolo_test.mp4 NULL c
+```
+The demo3D program takes the same parameters of the demo program:
+```
+./demo3D <network-rt-file> <path-to-video> <calibration-file> <kind-of-network> <number-of-classes> <n-batches> <show-flag> <conf-thresh>
+```
+where
+
+* ```<calibration-file>``` is the camera calibration file (opencv format). It is important that the file contains entry "camera_matrix" with sub-entry "rows", "cols", "data". If you do not want to pass the calibration file, pass "NULL" instead.
+
+![demo](https://user-images.githubusercontent.com/11939259/126784875-c4285497-d369-424f-abda-58274cd747ac.gif)
+
+## Object Detection and Tracking
+
+To run the 3D object detection & tracking demo follow these steps (example with CenterTrack based on DLA34):
+```
+rm dla34_ctrack_fp32.rt  # be sure to delete(or move) old tensorRT files
+./test_dla34_ctrack      # run the yolo test (is slow)
+./demoTracker dla34_ctrack_fp32.rt ../demo/yolo_test.mp4 NULL c
+```
+
+The demoTracker program takes the same parameters of the demo program:
+```
+./demoTracker <network-rt-file> <path-to-video> <calibration-file> <kind-of-network> <number-of-classes> <n-batches> <show-flag> <conf-thresh> <2D/3D-flag>
+```
+
+where
+
+* ```<calibration-file>``` is the camera calibration file (opencv format). It is important that the file contains entry "camera_matrix" with sub-entry "rows", "cols", "data". If you do not want to pass the calibration file, pass "NULL" instead.
+*  ```<2D/3D-flag>``` if set to 0 the demo will be in the 2D mode, while if set to 1 the demo will be in the 3D mode (Default is 1 - 3D mode).
+
+![demo](https://user-images.githubusercontent.com/11939259/126784878-513fa9e8-864a-4c24-b4bd-199737184708.gif)
+
+## FPS Results
+
+Inference FPS of shelfnet with tkDNN, average of 1200 images on:
+  * RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5);
+  * Xavier AGX, Jetpack 4.3 (CUDA 10.0, CUDNN 7.6.3, tensorrt 6.0.1 );
+
+### 3D OD and Tracking
+
+| Platform   | Test                        | Phase   | FP32, ms  | FP32, FPS | FP16, ms  |	FP16, FPS | INT8, ms  | INT8, FPS | 
+| :------:   | :-----:                     | :-----: | :-----:   | :-----:   | :-----:   |	:-----:   | :-----:   | :-----:   | 
+| RTX 2080Ti | CenterTrack3D 512x512 (B=1) | pre     | 4.43883   |  225.285  |  4.42951  |  225.759   |  4.44278  |  225.084  |
+| RTX 2080Ti | CenterTrack3D 512x512 (B=1) | inf     | 9.03454   |  110.686  |  6.02013  |  166.109   |  5.31611  |  188.108  |
+| RTX 2080Ti | CenterTrack3D 512x512 (B=1) | post    | 0.96631   |  1034.87  |  0.96824  |  1032.80   |  0.95066  |  1051.90  |
+| RTX 2080Ti | CenterTrack3D 512x512 (B=1) | tot     | 14.4397   |  69.2535  |  11.4179  |  87.5818   |  10.7095  |  93.3750  |
+| RTX 2080Ti | CenterTrack3D 512x512 (B=4) | pre     | 4.60075   |  217.356  |  4.28658  |  233.286   |  4.29473  |  232.844  | 
+| RTX 2080Ti | CenterTrack3D 512x512 (B=4) | inf     | 8.48365   |  117.874  |  5.25150  |  190.422   |  4.58463  |  218.120  |  
+| RTX 2080Ti | CenterTrack3D 512x512 (B=4) | post    | 0.99484   |  1005.19  |  0.91776  |  1089.61   |  0.89853  |  1112.93  |  
+| RTX 2080Ti | CenterTrack3D 512x512 (B=4) | tot     | 14.0792   |  71.0266  |  10.4558  |  95.6405   |  9.77788  |  102.272  |  
+| AGX Xavier | CenterTrack3D 512x512 (B=1) | pre     | 34.9915   |  28.5784  |  33.5976  |  29.7440   |  34.4425  |  29.0339  |
+| AGX Xavier | CenterTrack3D 512x512 (B=1) | inf     | 76.3579   |  13.0962  |  52.4759  |  19.0564   |  51.4610  |  19.4322  |
+| AGX Xavier | CenterTrack3D 512x512 (B=1) | post    | 3.38576   |  295.355  |  3.26010  |  306.739   |  3.19770  |  312.725  |
+| AGX Xavier | CenterTrack3D 512x512 (B=1) | tot     | 114.735   |  8.71574  |  89.3336  |  11.1940   |  89.1012  |  11.2232  |
+| AGX Xavier | CenterTrack3D 512x512 (B=4) | pre     | 32.8933   |  30.4014  |  32.7950  |  30.4925   |  32.9603  |  30.3396  | 
+| AGX Xavier | CenterTrack3D 512x512 (B=4) | inf     | 74.2840   |  13.4618  |  50.3858  |  19.8469   |  49.2030  |  20.3240  |  
+| AGX Xavier | CenterTrack3D 512x512 (B=4) | post    | 3.14888   |  317.574  |  3.13615  |  318.862   |  3.02550  |  330.524  |  
+| AGX Xavier | CenterTrack3D 512x512 (B=4) | tot     | 110.326   |  9.06404  |  86.3169  |  11.5852   |  85.1888  |  11.7386  |  
+
+
+### 2D OD and Tracking
+
+| Platform   | Test                        | Phase   | FP32, ms  | FP32, FPS | FP16, ms  |	FP16, FPS | INT8, ms  | INT8, FPS | 
+| :------:   | :-----:                     | :-----: | :-----:   | :-----:   | :-----:   |	:-----:   | :-----:   | :-----:   | 
+| RTX 2080Ti | CenterTrack2D 512x512 (B=1) | pre     | 4.44386   |  225.030  |  4.43828  |  225.313   |  4.47747  |  223.340  |
+| RTX 2080Ti | CenterTrack2D 512x512 (B=1) | inf     | 9.08365   |  110.088  |  6.04842  |  165.332   |  5.34787  |  186.990  |
+| RTX 2080Ti | CenterTrack2D 512x512 (B=1) | post    | 0.98593   |  1014.27  |  0.97745  |  1023.07   |  0.96595  |  1035.25  |
+| RTX 2080Ti | CenterTrack2D 512x512 (B=1) | tot     | 14.5134   |  68.9018  |  11.4642  |  87.2281   |  10.7913  |  92.6672  |
+| RTX 2080Ti | CenterTrack2D 512x512 (B=4) | pre     | 4.41188   |  226.661  |  4.50800  |  221.828   |  4.29238  |  232.971  | 
+| RTX 2080Ti | CenterTrack2D 512x512 (B=4) | inf     | 8.29015   |  120.625  |  5.38630  |  185.656   |  4.58500  |  218.103  |  
+| RTX 2080Ti | CenterTrack2D 512x512 (B=4) | post    | 0.96847   |  1032.55  |  0.97997  |  1020.44   |  0.91791  |  1089.43  |  
+| RTX 2080Ti | CenterTrack2D 512x512 (B=4) | tot     | 13.6705   |  73.1502  |  10.8743  |  91.9602   |  9.79528  |  102.090  |  
+| AGX Xavier | CenterTrack2D 512x512 (B=1) | pre     | 33.4745   |  29.8735  |  33.4847  |  29.8643   |  33.5022  |  29.8488  |
+| AGX Xavier | CenterTrack2D 512x512 (B=1) | inf     | 76.2077   |  13.1220  |  52.5111  |  19.0436   |  51.6057  |  19.3777  |
+| AGX Xavier | CenterTrack2D 512x512 (B=1) | post    | 3.26055   |  306.697  |  3.26806  |  305.992   |  3.21988  |  310.571  |
+| AGX Xavier | CenterTrack2D 512x512 (B=1) | tot     | 111.943   |  8.93312  |  89.2639  |  11.2027   |  88.3278  |  11.3215  |
+| AGX Xavier | CenterTrack2D 512x512 (B=4) | pre     | 32.8323   |  30.4579  |  32.8595  |  30.4326   |  32.8195  |  30.4697  | 
+| AGX Xavier | CenterTrack2D 512x512 (B=4) | inf     | 74.3075   |  13.4576  |  50.3555  |  19.8588   |  49.1805  |  20.3333  |  
+| AGX Xavier | CenterTrack2D 512x512 (B=4) | post    | 3.12360   |  320.143  |  3.13570  |  318.908   |  3.04943  |  327.931  |  
+| AGX Xavier | CenterTrack2D 512x512 (B=4) | tot     | 110.263   |  9.06920  |  86.3507  |  11.5807   |  85.0494  |  11.7579  |  
+

docs/README_depth.md

@@ -0,0 +1,54 @@
+# Monocular depth estimation with tkDNN
+
+Currently tkDNN supports only Monodepth2 as monocular depth esitmation network.
+
+
+## Run the demo
+
+To run the depth estimation demo follow these steps (example with monodepth2):
+```
+rm monodepth2_fp32.rt        # be sure to delete(or move) old tensorRT files
+./test_monodepth2            # run the yolo test (is slow)
+./demoDepth monodepth2_fp32.rt ../demo/yolo_test.mp4 
+```
+In general the demo program takes the following parameters:
+```
+./demoDepth <network-rt-file> <path-to-video> <show-flag> <save-flag>
+```
+where
+*  ```<network-rt-file>``` is the rt file generated by a test
+*  ```<<path-to-video>``` is the path to a video file or a camera input  
+*  ```<show-flag>``` if set to 0 the demo will not show the visualization, it will otherwise (default=1)
+*  ```<save-flag>``` if set to 1 the demo will save the video into result.mp4, it won't otherwise (default=1)
+
+NB) By default it is used FP32 inference
+
+
+![demo](https://user-images.githubusercontent.com/11939259/160845358-0d6ab15d-c5f4-46ae-b9da-bfaf3903389d.gif "Results on yolo_test.mp4")  
+
+
+<!-- ## FPS Results
+
+Inference FPS of shelfnet with tkDNN, average of 1200 images on:
+  * RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5);
+  * Xavier AGX, Jetpack 4.3 (CUDA 10.0, CUDNN 7.6.3, tensorrt 6.0.1 );
+
+| Platform   | Test                     | Phase   | FP32, ms  | FP32, FPS | FP16, ms  |	FP16, FPS  | INT8, ms |	INT8, FPS | 
+| :------:   | :-----:                  | :-----: | :-----:   | :-----:   | :-----:   |	:-----:    | :-----:  |	:-----:   | 
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | pre     | 6.11863   |  163.435  |   5.81465 |  171.979   |  5.88699 |   169.866 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | inf     | 11.5464   |  86.6074  |   7.35396 |  135.981   |  6.37623 |   156.832 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | post    | 4.09058   |  244.464  |   3.91961 |  255.128   |  4.07343 |   245.493 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | tot     | 21.7556   |  45.9652  |   17.0882 |  58.5199   |  16.3366 |   61.2121 |
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | pre     | 25.435    |  39.3158  |   25.2953 |  39.5331   |  25.9303 |   38.565  | 
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | inf     | 36.5015   |  27.3961  |   17.0534 |  58.6395   |  15.6061 |   64.0773 |  
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | post    | 17.3917   |  57.4985  |   17.1649 |  58.2583   |  17.5539 |   56.9675 |  
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | tot     | 79.3283   |  12.6058  |   59.5136 |  16.8029   |  59.0903 |   16.9233 |  
+| AGX Xavier | shelfnet 1024x1024 (B=1) | pre     | 8.0174    |  124.729  |   7.5117  |  133.126   |  7.47333 |   133.809 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | inf     | 72.4173   |  13.8089  |   37.505  |  26.6631   |  31.3286 |   31.9197 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | post    | 8.89958   |  112.365  |   8.83576 |  113.176   |  9.42655 |   106.083 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | tot     | 89.3342   |  11.1939  |   53.8525 |  18.5692   |  48.2285 |   20.7346 |
+| AGX Xavier | shelfnet 2048x2048 (B=4) | pre     | 47.1454   |  21.211   |   21.6475 |  46.1947   |  21.4201 |   46.6851 | 
+| AGX Xavier | shelfnet 2048x2048 (B=4) | inf     | 266.537   |  3.75183  |   128.321 |  7.79293   |  107.621 |   9.29185 |  
+| AGX Xavier | shelfnet 2048x2048 (B=4) | post    | 44.0711   |  22.6906  |   40.1732 |  24.8922   |  39.873  |   25.0796 |  
+| AGX Xavier | shelfnet 2048x2048 (B=4) | tot     | 357.753   |  2.79522  |   190.142 |  5.25922   |  168.914 |   5.92016 |   -->
+

docs/README_seg.md

@@ -0,0 +1,68 @@
+# Semantic Segmentation with tkDNN
+
+Currently tkDNN supports only ShelfNet as semantic segmentation network.
+
+
+## Run the demo
+
+To run the semantic segmentation demo follow these steps (example with shelfnet):
+```
+rm shelfnet_fp32.rt        # be sure to delete(or move) old tensorRT files
+export TKDNN_BATCHSIZE=4   # be sure you have batch size > than 1 if you want to run inference on images bigger than 1024
+./test_shelfnet            # run the yolo test (is slow)
+./demo shelfnet_fp32.rt ../demo/yolo_test.mp4 1 19
+```
+In general the demo program takes the following parameters:
+```
+./seg_demo <network-rt-file> <path-to-video> <n-batches> <number-of-classes> <resize-flag> <baseline-resize> <show-flag> <write-pred>
+```
+where
+*  ```<network-rt-file>``` is the rt file generated by a test
+*  ```<<path-to-video>``` is the path to a video file or a camera input  
+*  ```<n-batches>``` number of batches to use in inference (N.B. you should first export TKDNN_BATCHSIZE to the required n_batches and create again the rt file for the network).
+*  ```<number-of-classes>```is the number of classes the network is trained on
+*  ```<resize-flag>``` if set to 0 the demo will not resize the input frames, but use it as it is, otherwise it will resize it.
+*  ```<baseline-resize>``` is ```<resize-flag>``` is set to 1, then the input frames will be proportionally resized using ```<baseline-resize>``` as width baseline.
+*  ```<show-flag>``` if set to 0 the demo will not show the visualization but save the video into result.mp4 (if n-batches ==1)
+*  ```<write-pred>``` if set to 0 (default) the demo will run, otherwise the evaluation of a dataset will run and the output of the segmentation will be saved. Attention: this is under development and paths are embedded, so change them in the code in advance.
+
+NB) By default it is used FP32 inference
+NB) The batching is not used to work on more streams, rather to work on more tiles of the same image. Shelfnet never resized the input image, therefore for images greater than 1024x1024 tiles of 1024x1024 are given in input to the network in batch. 
+
+![demo](https://user-images.githubusercontent.com/11939259/126784236-38d24fc3-02df-4514-81c4-497e87e40b65.gif "Results on yolo_test.mp4")  
+
+For other demo videos refer to [this playlist](https://www.youtube.com/playlist?list=PLv0nEQYDD45y5EdSiywwCGPBmJVUzIWwe).
+
+NB) The gif and the videos are obtained with Mapillary Vistas weights, that we cannot publicly share due to its license restrictions. However, you can train Shelfnet using Mapillary and [this](https://git.hipert.unimore.it/mverucchi/shelfnet) fork of the original repo.
+
+
+## FPS Results
+
+Inference FPS of shelfnet with tkDNN, average of 1200 images on:
+  * RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5);
+  * Xavier AGX, Jetpack 4.3 (CUDA 10.0, CUDNN 7.6.3, tensorrt 6.0.1 );
+
+| Platform   | Test                     | Phase   | FP32, ms  | FP32, FPS | FP16, ms  |	FP16, FPS  | INT8, ms |	INT8, FPS | 
+| :------:   | :-----:                  | :-----: | :-----:   | :-----:   | :-----:   |	:-----:    | :-----:  |	:-----:   | 
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | pre     | 6.11863   |  163.435  |   5.81465 |  171.979   |  5.88699 |   169.866 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | inf     | 11.5464   |  86.6074  |   7.35396 |  135.981   |  6.37623 |   156.832 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | post    | 4.09058   |  244.464  |   3.91961 |  255.128   |  4.07343 |   245.493 |
+| RTX 2080Ti | shelfnet 1024x1024 (B=1) | tot     | 21.7556   |  45.9652  |   17.0882 |  58.5199   |  16.3366 |   61.2121 |
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | pre     | 25.435    |  39.3158  |   25.2953 |  39.5331   |  25.9303 |   38.565  | 
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | inf     | 36.5015   |  27.3961  |   17.0534 |  58.6395   |  15.6061 |   64.0773 |  
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | post    | 17.3917   |  57.4985  |   17.1649 |  58.2583   |  17.5539 |   56.9675 |  
+| RTX 2080Ti | shelfnet 2048x2048 (B=4) | tot     | 79.3283   |  12.6058  |   59.5136 |  16.8029   |  59.0903 |   16.9233 |  
+| AGX Xavier | shelfnet 1024x1024 (B=1) | pre     | 8.0174    |  124.729  |   7.5117  |  133.126   |  7.47333 |   133.809 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | inf     | 72.4173   |  13.8089  |   37.505  |  26.6631   |  31.3286 |   31.9197 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | post    | 8.89958   |  112.365  |   8.83576 |  113.176   |  9.42655 |   106.083 |
+| AGX Xavier | shelfnet 1024x1024 (B=1) | tot     | 89.3342   |  11.1939  |   53.8525 |  18.5692   |  48.2285 |   20.7346 |
+| AGX Xavier | shelfnet 2048x2048 (B=4) | pre     | 47.1454   |  21.211   |   21.6475 |  46.1947   |  21.4201 |   46.6851 | 
+| AGX Xavier | shelfnet 2048x2048 (B=4) | inf     | 266.537   |  3.75183  |   128.321 |  7.79293   |  107.621 |   9.29185 |  
+| AGX Xavier | shelfnet 2048x2048 (B=4) | post    | 44.0711   |  22.6906  |   40.1732 |  24.8922   |  39.873  |   25.0796 |  
+| AGX Xavier | shelfnet 2048x2048 (B=4) | tot     | 357.753   |  2.79522  |   190.142 |  5.25922   |  168.914 |   5.92016 |  
+
+
+## Known issues
+
+When creating the rt file all the checks returns errors. It is due to a different resize function and handling of the original ShelfNet outputs.
+However, the network is supposed to work. 

docs/demo.md

@@ -0,0 +1,120 @@
+# 2D Object Detection with tkDNN
+
+## Supported Networks
+
+* Yolo4, Yolo4-csp, Yolo4x, Yolo4_berkeley, Yolo4tiny
+* Yolo3, Yolo3_berkeley, Yolo3_coco4, Yolo3_flir, Yolo3_512, Yolo3tiny, Yolo3tiny_512
+* Yolo2, Yolo2_voc, Yolo2tiny
+* Csresnext50-panet-spp, Csresnext50-panet-spp_berkeley
+* Resnet101_cnet, Dla34_cnet
+* Mobilenetv2ssd, Mobilenetv2ssd512, Bdd-mobilenetv2ssd
+
+## Index
+
+ - [2D Object Detection](#2d-object-detection)
+ - [FP16 inference](#fp16-inference)
+ - [INT8 inference](#int8-inference)
+ - [Batching](#batching)
+
+### 2D Object Detection
+This is an example using yolov4.
+
+To run the an object detection first create the .rt file by running:
+```
+rm yolo4_fp32.rt        # be sure to delete(or move) old tensorRT files
+./test_yolo4            # run the yolo test (is slow)
+```
+If you get problems in the creation, try to check the error activating the debug of TensorRT in this way:
+```
+cmake .. -DCMAKE_BUILD_TYPE=Debug -DDEBUG=True
+make
+```
+
+Once you have successfully created your rt file, run the demo:
+```
+./ demo <path-to-config>
+```
+In general the demo program takes 1 parameter, the ```<path-to-config>``` that is the path to che configuration file. The parameter is optional and its default value is ```"../demo/demoConfig.yaml"```.
+
+The config file is a yaml file with the following attributes:
+* ```net``` is the rt file generated by a test
+* ```input``` is the path to a video file or a camera input  (on Linux)
+* ```win_input``` is the path to a video file or a camera input  (on Windows)
+* ```ntype``` is the type of network. Thee types are currently supported: ```y``` (YOLO family), ```c``` (CenterNet family) and ```m``` (MobileNet-SSD family)
+* ```n_classes``` is the number of classes the network is trained on
+* ```n_batch``` number of batches to use in inference (N.B. you should first export TKDNN_BATCHSIZE to the required n_batches and create again the rt file for the network).
+* ```conf_thresh``` confidence threshold for the detector. Only bounding boxes with threshold greater than conf-thresh will be displayed.
+* ```show``` if set to 0 the demo will not show the visualization (if n-batches ==1)
+* ```save``` if set to 1 the demo will save the video of the demo into result.mp4 (if n-batches ==1)
+
+N.B. By default it is used FP32 inference
+
+
+![demo](https://user-images.githubusercontent.com/11562617/72547657-540e7800-388d-11ea-83c6-49dfea2a0607.gif)
+
+
+### FP16 inference
+
+To run the demo with FP16 inference follow these steps (example with yolov3):
+```
+export TKDNN_MODE=FP16  # set the half floating point optimization
+rm yolo4_fp16.rt        # be sure to delete(or move) old tensorRT files
+./test_yolo4            # run the yolo test (is slow)
+#set net: yolo4_fp16.rt in the config file
+./demo
+```
+N.B. Using FP16 inference will lead to some errors in the results (first or second decimal). 
+
+### INT8 inference
+
+To run the demo with INT8 inference three environment variables need to be set:
+
+  * ```export TKDNN_MODE=INT8```: set the 8-bit integer optimization
+  * ```export TKDNN_CALIB_IMG_PATH=/path/to/calibration/image_list.txt``` : image_list.txt has in each line the absolute path to a calibration image
+  * ```export TKDNN_CALIB_LABEL_PATH=/path/to/calibration/label_list.txt```: label_list.txt has in each line the absolute path to a calibration label
+  
+You should provide image_list.txt and label_list.txt, using training images. However, if you want to quickly test the INT8 inference you can run (from this repo root folder)
+```
+bash scripts/download_validation.sh COCO
+```
+to automatically download COCO2017 validation (inside demo folder) and create those needed file. Use BDD instead of COCO to download BDD validation. 
+
+Then a complete example using yolo3 and COCO dataset would be:
+```
+export TKDNN_MODE=INT8
+export TKDNN_CALIB_LABEL_PATH=../demo/COCO_val2017/all_labels.txt
+export TKDNN_CALIB_IMG_PATH=../demo/COCO_val2017/all_images.txt
+rm yolo4_int8.rt        # be sure to delete(or move) old tensorRT files
+./test_yolo4            # run the yolo test (is slow)
+#set net: yolo4_int8.rt in the config file
+./demo
+```
+N.B. 
+
+ * Using INT8 inference will lead to some errors in the results. 
+ * The test will be slower: this is due to the INT8 calibration, which may take some time to complete. 
+ * INT8 calibration requires TensorRT version greater than or equal to 6.0
+ * Only 100 images are used to create the calibration table by default (set in the code).
+
+### Batching
+
+#### BatchSize bigger than 1
+```
+export TKDNN_BATCHSIZE=2
+# build tensorRT files
+```
+This will create a TensorRT file with the desired **max** batch size.
+The test will still run with a batch of 1, but the created tensorRT can manage the desired batch size.
+
+#### Test batch Inference
+This will test the network with random input and check if the output of each batch is the same.
+```
+./test_rtinference <network-rt-file> <number-of-batches>
+# <number-of-batches> should be less or equal to the max batch size of the <network-rt-file>
+
+# example
+export TKDNN_BATCHSIZE=4           # set max batch size
+rm yolo3_fp32.rt                   # be sure to delete(or move) old tensorRT files
+./test_yolo3                       # build RT file
+./test_rtinference yolo3_fp32.rt 4 # test with a batch size of 4
+```

docs/exporting_weights.md

@@ -0,0 +1,127 @@
+# tkDNN export weights
+
+## Index 
+
+ - [How to export weights](#how-to-export-weights)
+    - [1)Export weights from darknet](#1export-weights-from-darknet)
+    - [2)Export weights for DLA34 and ResNet101](#2export-weights-for-dla34-and-resnet101)
+    - [3)Export weights for CenterNet](#3export-weights-for-centernet)
+    - [4)Export weights for MobileNetSSD](#4export-weights-for-mobilenetssd)
+    - [5)Export weights for CenterTrack](#5export-weights-for-centertrack)
+    - [6)Export weights for ShelfNet](#6export-weights-for-shelfnet)
+ - [Darknet Parser](#darknet-parser)
+
+## How to export weights
+
+Weights are essential for any network to run inference. For each test a folder organized as follow is needed (in the build folder):
+```
+    test_nn
+        |---- layers/ (folder containing a binary file for each layer with the corresponding wieghts and bias)
+        |---- debug/  (folder containing a binary file for each layer with the corresponding outputs)
+```
+Therefore, once the weights have been exported, the folders layers and debug should be placed in the corresponding test.
+
+### 1)Export weights from darknet
+To export weights for NNs that are defined in darknet framework, use [this](https://git.hipert.unimore.it/fgatti/darknet.git) fork of darknet and follow these steps to obtain a correct debug and layers folder, ready for tkDNN.
+
+```
+git clone https://git.hipert.unimore.it/fgatti/darknet.git
+cd darknet
+make
+mkdir layers debug
+./darknet export <path-to-cfg-file> <path-to-weights> layers
+```
+N.B. Use compilation with CPU (leave GPU=0 in Makefile) if you also want debug. 
+
+### 2)Export weights for DLA34 and ResNet101 
+To get weights and outputs needed to run the tests dla34 and resnet101 use the Python script and the Anaconda environment included in the repository.   
+
+Create Anaconda environment and activate it:
+```
+conda env create -f file_name.yml
+source activate env_name
+python <script name>
+```
+### 3)Export weights for CenterNet
+To get the weights needed to run Centernet tests use [this](https://github.com/sapienzadavide/CenterNet.git) fork of the original Centernet. 
+```
+git clone https://github.com/sapienzadavide/CenterNet.git
+```
+* follow the instruction in the README.md and INSTALL.md
+
+```
+python demo.py --input_res 512 --arch resdcn_101 ctdet --demo /path/to/image/or/folder/or/video/or/webcam --load_model ../models/ctdet_coco_resdcn101.pth --exp_wo --exp_wo_dim 512
+python demo.py --input_res 512 --arch dla_34 ctdet --demo /path/to/image/or/folder/or/video/or/webcam --load_model ../models/ctdet_coco_dla_2x.pth --exp_wo --exp_wo_dim 512
+```
+### 4)Export weights for MobileNetSSD
+To get the weights needed to run Mobilenet tests use [this](https://github.com/mive93/pytorch-ssd) fork of a Pytorch implementation of SSD network. 
+
+```
+git clone https://github.com/mive93/pytorch-ssd
+cd pytorch-ssd
+conda env create -f env_mobv2ssd.yml
+python run_ssd_live_demo.py mb2-ssd-lite <pth-model-fil> <labels-file>
+```
+### 5)Export weights for CenterTrack
+To get the weights needed to run CenterTrack tests use [this](https://github.com/sapienzadavide/CenterTrack.git) fork of the original CenterTrack. 
+```
+git clone https://github.com/sapienzadavide/CenterTrack.git
+```
+* follow the instruction in the README.md and INSTALL.md
+
+```
+python demo.py tracking,ddd --load_model ../models/nuScenes_3Dtracking.pth --dataset nuscenes --pre_hm --track_thresh 0.1 --demo /path/to/image/or/folder/or/video/or/webcam --test_focal_length 633 --exp_wo --exp_wo_dim 512 --input_h 512 --input_w 512
+```
+
+### 6)Export weights for ShelfNet
+To get the weights needed to run Shelfnet tests use [this](https://git.hipert.unimore.it/mverucchi/shelfnet) fork of a Pytorch implementation of Shelfnet network. 
+
+```
+git clone https://git.hipert.unimore.it/mverucchi/shelfnet
+cd shelfnet 
+cd ShelfNet18_realtime
+conda env create --file shelfnet_env.yml
+conda activate shelfnet
+mkdir layer debug
+python export.py
+```
+
+### 6)Export weights for monodepth2
+To get the weights needed to run Shelfnet tests use [this](https://github.com/perseusdg/monodepth2) fork of a Pytorch implementation of monodepth2 network. 
+
+```
+git clone https://github.com/perseusdg/monodepth2
+cd monodepth2
+mkdir models # Download the official weights and put depth.pth and encorder.pth inside this new folder
+conda env create --file monodepth.yaml
+conda activate monodepth2
+python exporter.py # you will find the weights inside the tkDNN_bin folder
+```
+
+## Darknet Parser
+tkDNN implement and easy parser for darknet cfg files, a network can be converted with *tk::dnn::darknetParser*:
+```
+// example of parsing yolo4
+tk::dnn::Network *net = tk::dnn::darknetParser("yolov4.cfg", "yolov4/layers", "coco.names");
+net->print();
+```
+All models from darknet are now parsed directly from cfg, you still need to export the weights with the described tools in the previous section.
+<details>
+  <summary>Supported layers</summary>
+  convolutional
+  maxpool
+  avgpool
+  shortcut
+  upsample
+  route
+  reorg
+  region
+  yolo
+</details>
+<details>
+  <summary>Supported activations</summary>
+  relu
+  leaky
+  mish
+  logistic
+</details>

docs/mAP_demo.md

@@ -0,0 +1,32 @@
+# Run the mAP demo
+
+To compute mAP, precision, recall and f1score to evaluate 2D object detectors, run the map_demo.
+
+A validation set is needed. 
+To download COCO_val2017 (80 classes) run (form the root folder): 
+```
+bash scripts/download_validation.sh COCO
+```
+To download Berkeley_val (10 classes) run (form the root folder): 
+```
+bash scripts/download_validation.sh BDD
+```
+
+To compute the map, the following parameters are needed:
+```
+./map_demo <network rt> <network type [y|c|m]> <labels file path> <config file path>
+```
+where 
+* ```<network rt>```: rt file of a chosen network on which compute the mAP.
+* ```<network type [y|c|m]>```: type of network. Right now only y(yolo), c(centernet) and m(mobilenet) are allowed
+* ```<labels file path>```: path to a text file containing all the paths of the ground-truth labels. It is important that all the labels of the ground-truth are in a folder called 'labels'. In the folder containing the folder 'labels' there should be also a folder 'images', containing all the ground-truth images having the same same as the labels. To better understand, if there is a label path/to/labels/000001.txt there should be a corresponding image path/to/images/000001.jpg. 
+* ```<config file path>```: path to a yaml file with the parameters needed for the mAP computation, similar to demo/config.yaml
+
+Example:
+
+```
+cd build
+./map_demo dla34_cnet_FP32.rt c ../demo/COCO_val2017/all_labels.txt ../demo/config.yaml
+```
+
+This demo also creates a json file named ```net_name_COCO_res.json``` containing all the detections computed. The detections are in COCO format, the correct format to submit the results to [CodaLab COCO detection challenge](https://competitions.codalab.org/competitions/20794#participate).

docs/windows.md

@@ -0,0 +1,102 @@
+# tkDNN on Windows 
+
+## Index
+
+ - [Dependencies-Windows](#dependencies-windows)
+ - [Compiling tkDNN on Windows](#compiling-tkdnn-on-windows)
+ - [Run the demo on Windows](#run-the-demo-on-windows)
+    - [FP16 inference windows](#fp16-inference-windows)
+    - [INT8 inference windows](#int8-inference-windows)
+ - [Run tkDNN on WSL2 with cuda](#tkdnn-on-cuda-wsl)
+ - [Known issues with tkDNN on Windows](#known-issues-with-tkdnn-on-windows)
+
+### Dependencies-Windows 
+This branch should work on every NVIDIA GPU supported in windows with the following dependencies:
+
+* WINDOWS 10 1803/WINDOWS 11 or HIGHER 
+* CUDA 11.2 
+* CUDNN 8.1.1
+* TENSORRT 7.2.3 
+* OPENCV 4.2  
+* MSVC 16.9+
+* YAML-CPP 
+* EIGEN3
+* 7ZIP (ADD TO PATH)
+* NINJA 1.10
+
+
+All the above mentioned dependencies except 7ZIP can be installed using Microsoft's [VCPKG](https://github.com/microsoft/vcpkg.git) .
+After bootstrapping VCPKG the dependencies can be built and installed using the following command :
+
+```
+opencv4(normal) - vcpkg.exe install opencv4[tbb,jpeg,tiff,opengl,openmp,png,ffmpeg,eigen]:x64-windows yaml-cpp:x64-windows eigen3:x64-windows --x-install-root=C:\opt --x-buildtrees-root=C:\temp_vcpkg_build
+
+opencv4(cuda) - vcpkg.exe install opencv4[cuda,nonfree,contrib,eigen,tbb,jpeg,tiff,opengl,openmp,png,ffmpeg]:x64-windows yaml-cpp:x64-windows eigen3:x64-windows --x-install-root=C:\opt --x-buildtrees-root=C:\temp_vcpkg_build
+```
+To build opencv4 with cuda and cudnn version corresponding to your cuda version,vcpkg's cudnn portfile needs to be modified by adding ```$ENV{CUDA_PATH}```  at lines 16 and 17 in the portfile.cmake 
+
+After VCPKG finishes building and installing all the packages delete C:\temp_vcpkg_build and add C:\opt\x64-windows\bin and C:\opt\x64-windows\debug\bin to path 
+
+### Compiling tkDNN on Windows 
+
+tkDNN is built with cmake(3.15+) on windows along with ninja.Msbuild and NMake Makefiles are drastically slower when compiling the library compared to windows
+```
+git clone https://github.com/ceccocats/tkDNN.git
+cd tkdnn-windows
+mkdir build
+cd build
+cmake -DCMAKE_BUILD_TYPE=Release -G"Ninja" ..
+ninja -j4
+```
+
+### Run the demo on Windows 
+
+This example uses yolo4_tiny.\
+To run the object detection file create .rt file bu running:
+```
+.\test_yolo4tiny.exe
+```
+
+Once the rt file has been successfully create,run the demo using the following command:
+```
+.\demo.exe yolo4_fp32.rt ..\demo\yolo_test.mp4 y 80 ..\tests\darknet\cfg\yolo4.cfg ..\tests\darknet\names\cococ.names 
+```
+ For general info on more demo paramters,check Run the demo section on top 
+ To run the test_all_tests.sh on windows,use git bash or msys2 
+
+### FP16 inference windows 
+
+This is an untested feature on windows.To run the object detection demo with FP16 interference follow the below steps(example with yolo4tiny):
+```
+set TKDNN_MODE=FP16
+del /f yolo4tiny_fp16.rt
+.\test_yolo4tiny.exe
+.\demo.exe yolo4tiny_fp16.rt ..\demo\yolo_test.mp4
+```
+
+### INT8 inference windows 
+To run object detection demo with INT8 (example with yolo4tiny):
+```
+set TKDNN_MODE=INT8
+set TKDNN_CALIB_LABEL_PATH=..\demo\COCO_val2017\all_labels.txt
+set TKDNN_CALIB_IMG_PATH=..\demo\COCO_val2017\all_images.txt
+del /f  yolo4tiny_int8.rt        # be sure to delete(or move) old tensorRT files
+.\test_yolo4tiny.exe           # run the yolo test (is slow)
+.\demo.exe yolo4tiny_int8.rt ..\demo\yolo_test.mp4 y
+
+```
+
+### Run tkDNN on WSL2 with cuda
+tkDNN works on wsl2 with cuda,although not all networks (centernet,mobilenet) work properly.
+If you encounter issues with running the network as a result of driver not found or cuda launch error,running the following command should solve the issue
+```cp /usr/lib/wsl/lib/lib* /usr/lib/x86_64-linux-gnu/ ```
+
+
+
+### Known issues with tkDNN on Windows
+
+In theory all models (centernet,mobilenet,darknet,centertrack,cnet3d and shelfnet) should work on Windows.
+
+On pascal cards(sm 6x) ,nvidia cuda wsl driver 510.06 don't work well with tkDNN both on windows and cuda wsl , Nvidia drivers >465+ and < 500 are completely supported .
+
+

include/boxDetection.h

@@ -1,44 +0,0 @@
-#include <iostream>
-#include <sstream>
-#include <fstream>
-#include <iomanip>
-#include <stdlib.h>
-#include <cstring>
-#include <cstdlib>
-#include <time.h>
-#include <chrono>
-
-#include "cuda.h"
-#include "cuda_runtime_api.h"
-#include <cublas_v2.h>
-#include <cudnn.h>
-
-#include <opencv2/core/core.hpp>
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/imgproc/imgproc.hpp>
-
-//saliency
-#include <opencv2/core/utility.hpp>
-//#include <opencv2/saliency.hpp>
-#include <opencv2/highgui.hpp>
-
-#define SAVE false
-#define SAVE_TO(name, fn, i, var) {sprintf(buf_frame_crop_name,name,fn,i);\
-    cv::imwrite(buf_frame_crop_name, var);}
-
-
-
-// cv::Mat img_threshold(cv::Mat frame_crop);
-// cv::Mat img_background(cv::Mat frame_crop);
-// cv::Mat img_dist_transform(cv::Mat frame_crop);
-// cv::Mat img_watershed(cv::Mat frame_crop);
-void image_segmentation(cv::Mat frame_crop, int frame_nbr, int i);
-void image_gradients(cv::Mat frame_crop, int frame_nbr, int i);
-void image_find_contours(cv::Mat frame_crop, int frame_nbr, int i);
-void image_saliency(cv::Mat frame_crop, int frame_nbr, int i);
-void frame_box_disparity(cv::Mat pre_frame, cv::Mat frame, std::vector <cv::Rect>  pre_rois, int frame_nbr);
-void segmentation(cv::Mat pre_frame, cv::Mat frame_crop, int frame_nbr, int i, int mode);
-
-//canny
-cv::Mat img_laplacian(cv::Mat frame_crop, int ret);
-cv::Mat frame_disparity(cv::Mat pre_frame, cv::Mat frame, int frame_nbr, int i, int ret);

include/calibration.h

@@ -1,32 +0,0 @@
-#ifndef CALIBRATION_H
-#define CALIBRATION_H
-
-#include "gdal.h"
-#include <gdal_priv.h>
-#include <gdal/gdal.h>
-#include "gdal/gdal_priv.h"
-#include "gdal/cpl_conv.h"
-
-#include <yaml-cpp/yaml.h>
-#include <opencv2/calib3d.hpp>
-#include <opencv2/core.hpp>
-
-#include <iostream>
-#include <cstring>
-
-struct ObjCoords
-{
-    double lat_;
-    double long_;
-    int class_;
-};
-
-void readTiff(char *filename, double *adfGeoTransform);
-void readCameraCalibrationYaml(const std::string &cameraCalib, cv::Mat &cameraMat, cv::Mat &distCoeff);
-void pixel2coord(int x, int y, double &lat, double &lon, double *adfGeoTransform);
-void coord2pixel(double lat, double lon, int &x, int &y, double *adfGeoTransform);
-void fillMatrix(cv::Mat &H, double *matrix, bool show = false);
-void read_projection_matrix(cv::Mat &H, char *path);
-void convert_coords(std::vector<ObjCoords> &coords, int x, int y, int detected_class, cv::Mat H, double *adfGeoTransform);
-
-#endif /*CALIBRATION_H*/

include/cameraUtils.h

@@ -1,45 +0,0 @@
-#ifndef CAMERAUTILS_H
-#define CAMERAUTILS_H
-
-#include <vector>
-#include <mutex>
-#include <opencv2/core/core.hpp>
-#include "tracker.h"
-#include "Yolo3Detection.h"
-
-struct Camera_t
-{
-    int CAM_IDX;
-    char *input;
-    char *pmatrix;
-    char *maskfile;
-    char *cameraCalib;
-    char *maskFileOrient;
-    bool to_show;
-    tk::dnn::Yolo3Detection *yolo;
-    double adfGeoTransform[6];
-};
-
-struct Frame_t
-{
-    char *input;
-    cv::Mat frame;
-    int frame_nbr;
-    // sem_vc for mainthread, videocapturethread, originalthread and disparitythread
-    std::mutex sem_vc;
-};
-
-struct ModFrame_t
-{
-    std::vector<Tracker> trackers;
-    geodetic_converter::GeodeticConverter gc;
-    double adfGeoTransform[6];
-    cv::Mat H;
-    cv::Mat original_frame;
-    tk::dnn::Yolo3Detection yolo;
-    cv::Mat mask;
-    // sem for mainthread, detectionthread and topviewthread
-    std::mutex sem;
-};
-
-#endif /*CAMERAUTILS_H*/