Merge with master, all tests passed

Signed-off-by: Micaela Verucchi <micaelaverucchi@gmail.com>
2021-07-20 12:48:41 +02:00
parent 4ca69836e9 c306b36860
commit 84ff978ccb
73 changed files with 5207 additions and 305 deletions
@@ -12,5 +12,11 @@ build/
 *.hdf5
 *.pk
 *.table
+cmake-build-release/
 demo/COCO_val2017
-demo/BDD100K_val
+demo/BDD100K_val
+/.vs
+cmake-build-minsizerel/*
+scripts/COCO_val2017/*
+scripts/COCO_val2017.zip
+scripts/all_labels.txt
@@ -1,8 +1,15 @@
-cmake_minimum_required(VERSION 3.5)
+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 -Wno-deprecated-declarations -Wno-unused-variable")
+if(UNIX)
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fPIC -Wno-deprecated-declarations -Wno-unused-variable ")
+endif()
+if(WIN32)
+set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_FLAGS "/O2 /FS /EHsc")
+set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
+endif(WIN32)
 include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include/tkDNN)

 # project specific flags
@@ -10,7 +17,13 @@ if(DEBUG)
    add_definitions(-DDEBUG)
 endif()

-add_definitions(-DTKDNN_PATH="${CMAKE_CURRENT_SOURCE_DIR}")
+if(TKDNN_PATH)
+    message("SET TKDNN_PATH:"${TKDNN_PATH})
+    add_definitions(-DTKDNN_PATH="${TKDNN_PATH}")
+else()
+    add_definitions(-DTKDNN_PATH="${CMAKE_CURRENT_SOURCE_DIR}")
+endif()
+

 #-------------------------------------------------------------------------------
 # CUDA
@@ -28,19 +41,21 @@ include_directories(${CUDNN_INCLUDE_DIR})
 file(GLOB tkdnn_CUSRC "src/kernels/*.cu" "src/sorting.cu")
 cuda_include_directories(${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})


 #-------------------------------------------------------------------------------
 # 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")

 # gives problems in cross-compiling, probably malformed cmake config
-#find_package(yaml-cpp REQUIRED)
+find_package(yaml-cpp REQUIRED)

 #-------------------------------------------------------------------------------
 # Build Libraries
@@ -48,7 +63,7 @@ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DOPENCV")
 file(GLOB tkdnn_SRC "src/*.cpp")
 set(tkdnn_LIBS kernels ${CUDA_LIBRARIES} ${CUDA_CUBLAS_LIBRARIES} ${CUDNN_LIBRARIES} ${OpenCV_LIBS} yaml-cpp)

-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
+set(CMAKE_CXX_FLAGS "${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})
@@ -77,6 +92,7 @@ foreach(test_SRC ${darknet_SRC})
    set(test_NAME test_${test_NAME})
    add_executable(${test_NAME} ${test_SRC})
    target_link_libraries(${test_NAME} tkDNN)
+    install(TARGETS ${test_NAME} DESTINATION bin)
 endforeach()

 # MOBILENET
@@ -136,7 +152,10 @@ target_link_libraries(seg_demo tkDNN)
 message("install dir:" ${CMAKE_INSTALL_PREFIX})
 install(DIRECTORY include/ DESTINATION include/)
 install(TARGETS tkDNN kernels DESTINATION lib)
+install(TARGETS test_simple test_mnist test_mnistRT test_rtinference demo map_demo DESTINATION bin)
 install(DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/cmake/" # source directory
        DESTINATION "share/tkDNN/cmake/" # target directory
 )
-
+install(DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/tests/" # source directory
+        DESTINATION "share/tkDNN/tests" # target directory
+)
@@ -0,0 +1 @@
+1)error C2131 @ Yolo3Detection.cpp(97) -> expression doesnt evaluate to a constant caused to read of variable outside its lifetime
@@ -1,44 +1,70 @@
 # tkDNN
-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 and several discrete GPU.
+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. 


-If you use tkDNN in your research, please cite one of the following papers. For use in commercial solutions, write at gattifrancesco@hotmail.it or refer to https://hipert.unimore.it/ .
+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/ .

 ```
-Accepted paper @ IRC 2020, will soon be published.
-M. Verucchi, L. Bartoli, F. Bagni, F. Gatti, P. Burgio and M. Bertogna, "Real-Time clustering and LiDAR-camera fusion on embedded platforms for self-driving cars",  in proceedings in IEEE Robotic Computing (2020)
-
-Accepted paper @ ETFA 2020, will soon be published.
-M. Verucchi, G. Brilli, D. Sapienza, M. Verasani, M. Arena, F. Gatti, A. Capotondi, R. Cavicchioli, M. Bertogna, M. Solieri
-"A Systematic Assessment of Embedded Neural Networks for Object Detection", in IEEE International Conference on Emerging Technologies and Factory Automation (2020)
+@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}
+}
 ```

-## Results
-Inference FPS of yolov4 with tkDNN, average of 1200 images with the same dimesion as the input size, on 
+### What's new (20 July 2021)
+- [x] Support to sematic segmentation [REAME](readme/README_seg.md)
+- [] Support to TensorRT8 (WIP)
+
+## 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     |
-| 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      | -         | -         |
+| 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)
@@ -58,6 +84,14 @@ Inference FPS of yolov4 with tkDNN, average of 1200 images with the same dimesio
  - [mAP demo](#map-demo)
  - [Existing tests and supported networks](#existing-tests-and-supported-networks)
  - [References](#references)
+  - [tkDNN on Windows 10 (experimental)](#tkdnn-on-windows-10-experimental)
+    - [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)
+    - [Known issues with tkDNN on Windows](#known-issues-with-tkdnn-on-windows)
+  



@@ -155,7 +189,7 @@ tkDNN implement and easy parser for darknet cfg files, a network can be converte
 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 descripted tools in the previus section.
+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
@@ -173,19 +207,30 @@ All models from darknet are now parsed directly from cfg, you still need to expo
  relu
  leaky
  mish
+  logistic
 </details>

-## Run the demo
+## Run the demo 
+This is an example using yolov4.

-To run the an object detection demo follow these steps (example with yolov3):
+To run the an object detection first create the .rt file by running:
 ```
-rm yolo3_fp32.rt        # be sure to delete(or move) old tensorRT files
-./test_yolo3            # run the yolo test (is slow)
-./demo yolo3_fp32.rt ../demo/yolo_test.mp4 y
+rm yolo4_fp32.rt        # be sure to delete(or move) old tensorRT files
+./test_yolo4            # run the yolo test (is slow)
 ```
-In general the demo program takes 4 parameters:
+If you get problems in the creation, try to check the error activating the debug of TensorRT in this way:
 ```
-./demo <network-rt-file> <path-to-video> <kind-of-network> <number-of-classes> <n-batches> <show-flag>
+cmake .. -DDEBUG=True
+make
+```
+
+Once you have successfully created your rt file, run the demo: 
+```
+./demo yolo4_fp32.rt ../demo/yolo_test.mp4 y
+```
+In general the demo program takes 7 parameters:
+```
+./demo <network-rt-file> <path-to-video> <kind-of-network> <number-of-classes> <n-batches> <show-flag> <conf-thresh>
 ```
 where
 *  ```<network-rt-file>``` is the rt file generated by a test
@@ -194,9 +239,11 @@ where
 *  ```<number-of-classes>```is the number of classes the network is trained on
 *  ```<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).
 *  ```<show-flag>``` if set to 0 the demo will not show the visualization but save the video into result.mp4 (if n-batches ==1)
+*  ```<conf-thresh>``` confidence threshold for the detector. Only bounding boxes with threshold greater than conf-thresh will be displayed.

 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
@@ -221,7 +268,7 @@ You should provide image_list.txt and label_list.txt, using training images. How
 ```
 bash scripts/download_validation.sh COCO
 ```
-to automatically download COCO2017 validation (inside demo folder) and create those needed file. Use BDD insted of COCO to download BDD validation. 
+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:
 ```
@@ -243,8 +290,8 @@ N.B.
 export TKDNN_BATCHSIZE=2
 # build tensorRT files
 ```
-This will create a TensorRT file with the desidered **max** batch size.
-The test will still run with a batch of 1, but the created tensorRT can manage the desidered batch size.
+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.
@@ -290,7 +337,7 @@ 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 subit the results to [CodaLab COCO detection challenge](https://competitions.codalab.org/competitions/20794#participate).
+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).

 ## Existing tests and supported networks

@@ -317,6 +364,98 @@ This demo also creates a json file named ```net_name_COCO_res.json``` containing
 | 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_berkeley             | Yolov4 <sup>8</sup>                           | [BDD100K  ](https://bair.berkeley.edu/blog/2018/05/30/bdd/)                          | 10        | 540x320       | [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)     |
+| yolo4x-cps            | Scaled Yolov4 <sup>10</sup>                          | [COCO 2017](http://cocodataset.org/)                          | 80        | 512x512       | [weights](https://cloud.hipert.unimore.it/s/AfzHE4BfTeEm2gH/download)     |
+
+### tkDNN on Windows 10 (experimental)
+
+### Dependencies-Windows 
+This branch should work on every NVIDIA GPU supported in windows with the following dependencies:
+
+* WINDOWS 10 1803 or HIGHER 
+* CUDA 10.0 (Recommended CUDA 11.2 )
+* CUDNN 7.6 (Recommended CUDNN 8.1.1 )
+* TENSORRT 6.0.1 (Recommended TENSORRT 7.2.3.4 )
+* OPENCV 3.4 (Recommended OPENCV 4.2.0 )
+* MSVC 16.7 
+* 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 yolo4tiny_fp32.rt ..\demo\yolo_test.mp4 y 
+```
+ 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
+
+```
+
+### Known issues with tkDNN on Windows
+
+Mobilenet and Centernet demos work properly only when built with msvc 16.7 in Release Mode,when built in debug mode for the mentioned networks one might encounter opencv assert errors
+
+All Darknet models work properly with demo using MSVC version(16.7-16.9)
+
+It is recommended to use Nvidia Driver(465+),Cuda unknown errors have been observed when using older drivers on pascal(SM 61) devices.
+
+


 ## References
@@ -329,3 +468,5 @@ This demo also creates a json file named ```net_name_COCO_res.json``` containing
 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).
@@ -3,5 +3,5 @@ map_points  : 101   #number of recall points (0 for all, 101 for COCO, 11 Pascal
 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
+conf_thresh : 0.001   #threshold on the condifence of the bbox
 verbose     : false #print on screen information
@@ -1,7 +1,7 @@
 #include <iostream>
 #include <signal.h>
 #include <stdlib.h>     /* srand, rand */
-#include <unistd.h>
+//#include <unistd.h>
 #include <mutex>

 #include "CenternetDetection.h"
@@ -22,10 +22,15 @@ int main(int argc, char *argv[]) {
    signal(SIGINT, sig_handler);


-    std::string net = "yolo3_berkeley.rt";
+    std::string net = "yolo4tiny_fp32.rt";
    if(argc > 1)
        net = argv[1]; 
-    std::string input = "../demo/yolo_test.mp4";
+    #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]; 
    char ntype = 'y';
@@ -40,6 +45,9 @@ int main(int argc, char *argv[]) {
    bool show = true;
    if(argc > 6)
        show = atoi(argv[6]); 
+    float conf_thresh=0.3;
+    if(argc > 7)
+        conf_thresh = atof(argv[7]);     

    if(n_batch < 1 || n_batch > 64)
        FatalError("Batch dim not supported");
@@ -69,7 +77,7 @@ int main(int argc, char *argv[]) {
        FatalError("Network type not allowed (3rd parameter)\n");
    }

-    detNN->init(net, n_classes, n_batch);
+    detNN->init(net, n_classes, n_batch, conf_thresh);

    gRun = true;

@@ -128,7 +136,7 @@ int main(int argc, char *argv[]) {
    double mean = 0; 
    
    std::cout<<COL_GREENB<<"\n\nTime stats:\n";
-    std::cout<<"Min: "<<*std::min_element(detNN->stats.begin(), detNN->stats.end())/n_batch<<" ms\n";    
+   std::cout<<"Min: "<<*std::min_element(detNN->stats.begin(), detNN->stats.end())/n_batch<<" ms\n";    
    std::cout<<"Max: "<<*std::max_element(detNN->stats.begin(), detNN->stats.end())/n_batch<<" ms\n";    
    for(int i=0; i<detNN->stats.size(); i++) mean += detNN->stats[i]; mean /= detNN->stats.size();
    std::cout<<"Avg: "<<mean/n_batch<<" ms\t"<<1000/(mean/n_batch)<<" FPS\n"<<COL_END;   
@@ -2,7 +2,10 @@
 #include <iostream>
 #include <signal.h>
 #include <stdlib.h>     /* srand, rand */
+#ifdef __linux__
 #include <unistd.h>
+#endif
+
 #include <mutex>
 #include "utils.h"

@@ -105,7 +108,7 @@ int main(int argc, char *argv[])
        default:
            FatalError("Network type not allowed (3rd parameter)\n");
    }
-    detNN->init(net, n_classes);
+    detNN->init(net, n_classes, 1, conf_thresh);

    //read images 
    std::ifstream all_labels(labels_path);
@@ -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
+
+
@@ -0,0 +1,57 @@
+FROM nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04
+LABEL maintainer "Francesco Gatti"
+
+ADD nv-tensorrt-repo-ubuntu1804-cuda10.2-trt7.0.0.11-ga-20191216_1-1_amd64.deb /tmp/trt.deb
+RUN apt-get update && dpkg -i /tmp/trt.deb && rm /tmp/trt.deb && apt-get update
+RUN apt install -y libnvinfer7=7.0.0-1+cuda10.2 libnvinfer-dev=7.0.0-1+cuda10.2
+RUN DEBIAN_FRONTEND=noninteractive apt install -y git wget libeigen3-dev libyaml-cpp-dev
+RUN cd /tmp && \
+    wget https://github.com/Kitware/CMake/releases/download/v3.17.3/cmake-3.17.3-Linux-x86_64.sh && \
+    chmod +x cmake-3.17.3-Linux-x86_64.sh && \
+    ./cmake-3.17.3-Linux-x86_64.sh --prefix=/usr/local --exclude-subdir --skip-license && \
+    rm ./cmake-3.17.3-Linux-x86_64.sh
+
+RUN echo "INSTALL OPENCV"
+RUN apt-get install -y build-essential \
+    unzip \
+    pkg-config \
+    libjpeg-dev \
+    libpng-dev \
+    libtiff-dev \
+    libavcodec-dev \
+    libavformat-dev \
+    libswscale-dev \
+    libv4l-dev \
+    libxvidcore-dev \
+    libx264-dev \
+    libgtk-3-dev \
+    libatlas-base-dev \
+    gfortran \
+    libgstreamer1.0-dev \
+    libgstreamer-plugins-base1.0-dev \
+    libdc1394-22-dev \
+    libavresample-dev
+RUN cd && wget https://github.com/opencv/opencv/archive/4.3.0.tar.gz && tar -xf 4.3.0.tar.gz && rm *.tar.gz
+RUN cd && wget https://github.com/opencv/opencv_contrib/archive/4.3.0.tar.gz && tar -xf 4.3.0.tar.gz && rm *.tar.gz
+RUN cd && \
+    cd opencv-4.3.0 && 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='~/opencv_contrib-4.3.0/modules' \
+    -D BUILD_EXAMPLES=OFF \
+    -D WITH_CUDA=ON \
+    -D CUDA_ARCH_BIN=7.2 \
+    -D CUDA_ARCH_PTX="" \
+    -D ENABLE_FAST_MATH=ON \
+    -D CUDA_FAST_MATH=ON \
+    -D WITH_CUBLAS=ON \
+    -D WITH_LIBV4L=ON \
+    -D WITH_GSTREAMER=ON \
+    -D WITH_GSTREAMER_0_10=OFF \
+    -D WITH_TBB=ON \
+    ../ && make -j12 && make install
+RUN apt clean
+
+
@@ -0,0 +1,21 @@
+# 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
+
+# dowload tensorrt
+# from: https://developer.nvidia.com/compute/machine-learning/tensorrt/secure/7.0/7.0.0.11/local_repo/nv-tensorrt-repo-ubuntu1804-cuda10.2-trt7.0.0.11-ga-20191216_1-1_amd64.deb
+
+# build image
+docker build -t ceccocats/tkdnn:latest -f Dockerfile.base .
+
+# run image
+docker run -ti --gpus all --rm ceccocats/tkdnn:latest bash
+```
+
@@ -73,7 +73,7 @@ public:
    CenternetDetection() {};
    ~CenternetDetection() {}; 

-    bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1);
+    bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1, const float conf_thresh=0.3);
    void preprocess(cv::Mat &frame, const int bi=0);
    void postprocess(const int bi=0,const bool mAP=false);
 };
@@ -11,6 +11,7 @@ namespace tk { namespace dnn {
        int channels = 3;
        int batch_normalize=0;
        int groups = 1;
+        int group_id = 0;
        int filters=1;
        int size_x=1;
        int size_y=1;
@@ -23,7 +24,10 @@ namespace tk { namespace dnn {
        int num = 1;
        int pad = 0;
        int coords = 4;
+        int nms_kind = 0;
+        int new_coords= 0;
        float scale_xy = 1;
+        float nms_thresh = 0.45;
        std::vector<int> layers;
        std::string activation = "linear";

@@ -4,7 +4,10 @@
 #include <iostream>
 #include <signal.h>
 #include <stdlib.h>    
+#ifdef __linux__
 #include <unistd.h>
+#endif 
+
 #include <mutex>
 #include "utils.h"

@@ -14,7 +17,7 @@

 #include "tkdnn.h"

-// #define OPENCV_CUDACONTRIB //if OPENCV has been compiled with CUDA and contrib.
+//#define OPENCV_CUDACONTRIB //if OPENCV has been compiled with CUDA and contrib.

 #ifdef OPENCV_CUDACONTRIB
 #include <opencv2/cudawarping.hpp>
@@ -76,15 +79,15 @@ class DetectionNN {
        ~DetectionNN(){};

        /**
-         * Method used to inialize the class, allocate memory and compute 
+         * Method used to initialize the class, allocate memory and compute 
         * needed data.
         * 
-         * @param tensor_path path to the rt file og the NN.
+         * @param tensor_path path to the rt file of the NN.
         * @param n_classes number of classes for the given dataset.
         * @param n_batches maximum number of batches to use in inference
         * @return true if everything is correct, false otherwise.
         */
-        virtual bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1) = 0;
+        virtual bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1, const float conf_thresh=0.3) = 0;
        
        /**
         * This method performs the whole detection of the NN.
@@ -141,16 +144,15 @@ class DetectionNN {
        }      

        /**
-         * Method to draw boundixg boxes and labels on a frame.
+         * Method to draw bounding boxes and labels on a frame.
         * 
-         * @param frames orginal frame to draw bounding box on.
+         * @param frames original frame to draw bounding box on.
         */
        void draw(std::vector<cv::Mat>& frames) {
            tk::dnn::box b;
            int x0, w, x1, y0, h, y1;
            int objClass;
            std::string det_class;
-
            int baseline = 0;
            float font_scale = 0.5;
            int thickness = 2;   
@@ -1,7 +1,14 @@
 #include <iostream>
 #include <signal.h>
 #include <stdlib.h>     /* srand, rand */
+
+#ifdef __linux__
 #include <unistd.h>
+#elif _WIN32
+#define _USE_MATH_DEFINES
+#include <math.h>
+#endif 
+
 #include <mutex>
 #include <Eigen/Dense>
 #include "utils.h"
@@ -44,7 +51,7 @@ class ImuOdom {
        virtual ~ImuOdom() {}

        /**
-         * Method used for inizialize the class
+         * Method used for initialize the class
         * 
         * @return Success of the initialization
         */
@@ -141,7 +148,7 @@ class ImuOdom {
            //odomPOS = odomPOS + deltaP.cast<double>(); // V2
            odomROT = odomROT * q.normalized().toRotationMatrix();
            
-            // compute euler
+            // compute Euler
            auto newEULER = odomROT.eulerAngles(0, 1, 2);
            for(int i=0; i<3; i++) {
                while( fabs(newEULER(i) - odomEULER(i)) > M_PI_2 )  {
@@ -11,8 +11,11 @@
 #include <fstream>
 #include <iomanip>
 #include <signal.h>
-#include <stdlib.h>    
+#include <stdlib.h>
+#ifdef __linux__    
 #include <unistd.h>
+#endif
+
 #include <mutex>

 #include "NvInfer.h"
@@ -19,6 +19,7 @@ enum layerType_t {
    LAYER_ACTIVATION_CRELU,
    LAYER_ACTIVATION_LEAKY,
    LAYER_ACTIVATION_MISH,
+    LAYER_ACTIVATION_LOGISTIC,
    LAYER_FLATTEN,
    LAYER_RESHAPE,
    LAYER_RESIZE,
@@ -69,6 +70,7 @@ public:
            case LAYER_ACTIVATION_CRELU:    return "ActivationCReLU";
            case LAYER_ACTIVATION_LEAKY:    return "ActivationLeaky";
            case LAYER_ACTIVATION_MISH:     return "ActivationMish";
+            case LAYER_ACTIVATION_LOGISTIC: return "ActivationLogistic";
            case LAYER_FLATTEN:             return "Flatten";
            case LAYER_RESHAPE:             return "Reshape";
            case LAYER_RESIZE:              return "Resize";
@@ -173,7 +175,7 @@ public:


 /**
-    Input layer (it doesnt need weigths)
+    Input layer (it doesn't need weights)
 */
 class Input : public Layer {

@@ -209,16 +211,17 @@ public:


 /**
-    Avaible activation functions
+    Available activation functions
 */
 typedef enum {
    ACTIVATION_ELU     = 100,
    ACTIVATION_LEAKY   = 101,
-    ACTIVATION_MISH   = 102
+    ACTIVATION_MISH   = 102,
+    ACTIVATION_LOGISTIC   = 103
 } tkdnnActivationMode_t;

 /**
-    Activation layer (it doesnt need weigths)
+    Activation layer (it doesn't need weights)
 */
 class Activation : public Layer {

@@ -236,6 +239,8 @@ public:
            return LAYER_ACTIVATION_LEAKY;
        else if (act_mode == ACTIVATION_MISH)
            return LAYER_ACTIVATION_MISH;
+        else if (act_mode == ACTIVATION_LOGISTIC)
+            return LAYER_ACTIVATION_LOGISTIC;
        else
            return LAYER_ACTIVATION;
         };
@@ -276,8 +281,8 @@ public:
 protected:
    cudnnFilterDescriptor_t filterDesc;
    cudnnConvolutionDescriptor_t convDesc;
-    cudnnConvolutionFwdAlgo_t     algo;
-    cudnnConvolutionBwdDataAlgo_t bwAlgo;
+    cudnnConvolutionFwdAlgoPerf_t     algo;
+    cudnnConvolutionBwdDataAlgoPerf_t bwAlgo;
    cudnnTensorDescriptor_t biasTensorDesc;

    void initCUDNN(bool back = false);
@@ -321,9 +326,9 @@ public:
    virtual dnnType* infer(dataDim_t &dim, dnnType* srcData);

    const bool bidirectional = true; /**> is the net bidir */
-    bool returnSeq = false;       /**> if false return only the result of last timestep */
+    bool returnSeq = false;       /**> if false return only the result of last timestamp */
    int stateSize = 0; /**> number of hidden states */
-    int seqLen = 0;    /**> number of timesteps */
+    int seqLen = 0;    /**> number of timestamp */
    int numLayers = 1; /**> number of internal layers */

 protected:
@@ -370,7 +375,7 @@ public:


 /**
-    Deformable Convolutionl 2d layer
+    Deformable Convolutional 2d layer
 */  
 class DeformConv2d : public LayerWgs {

@@ -469,7 +474,7 @@ protected:


 /**
-    Avaible pooling functions (padding on tkDNN is not supported)
+    Available pooling functions (padding on tkDNN is not supported)
 */
 typedef enum {
    POOLING_MAX     = 0,
@@ -480,7 +485,7 @@ typedef enum {

 /**
    Pooling layer
-    currenty supported only 2d pooing (also on 3d input)
+    currently supported only 2d pooing (also on 3d input)
 */
 class Pooling : public Layer {

@@ -529,7 +534,7 @@ public:
 class Route : public Layer {

 public:
-    Route(Network *net, Layer **layers, int layers_n); 
+    Route(Network *net, Layer **layers, int layers_n, int groups = 1, int group_id = 0); 
    virtual ~Route();
    virtual layerType_t getLayerType() { return LAYER_ROUTE; };

@@ -539,12 +544,14 @@ public:
    static const int MAX_LAYERS = 32;
    Layer *layers[MAX_LAYERS];  //ids of layers to be merged
    int layers_n; //number of layers
+    int groups;
+    int group_id;
 };


 /**
    Reorg layer
-    Mantain same dimension but change C*H*W distribution
+    Maintains same dimension but change C*H*W distribution
 */
 class Reorg : public Layer {

@@ -578,7 +585,7 @@ public:

 /**
    Upsample layer
-    Mantain same dimension but change C*H*W distribution
+    Maintains same dimension but change C*H*W distribution
 */
 class Upsample : public Layer {

@@ -629,24 +636,28 @@ public:
        int sort_class;
    };

-    Yolo(Network *net, int classes, int num, std::string fname_weights,int n_masks=3, float scale_xy=1);
+    enum nmsKind_t {GREEDY_NMS=0, DIOU_NMS=1};
+
+    Yolo(Network *net, int classes, int num, std::string fname_weights,int n_masks=3, float scale_xy=1, double nms_thresh=0.45, nmsKind_t nsm_kind=GREEDY_NMS, int new_coords=0);
    virtual ~Yolo();
    virtual layerType_t getLayerType() { return LAYER_YOLO; };

-    int classes, num, n_masks;
+    int classes, num, n_masks, new_coords;
    dnnType *mask_h, *mask_d; //anchors
    dnnType *bias_h, *bias_d; //anchors
    float scaleXY;
+    double nms_thresh;
+    nmsKind_t nsm_kind; 
    std::vector<std::string> classesNames;

    virtual dnnType* infer(dataDim_t &dim, dnnType* srcData);
-    int computeDetections(Yolo::detection *dets, int &ndets, int netw, int neth, float thresh);
+    int computeDetections(Yolo::detection *dets, int &ndets, int netw, int neth, float thresh, int new_coords=0);

    dnnType *predictions;

-    static const int MAX_DETECTIONS = 8192;
+    static const int MAX_DETECTIONS = 8192*2;
    static Yolo::detection *allocateDetections(int nboxes, int classes);
-    static void             mergeDetections(Yolo::detection *dets, int ndets, int classes);
+    static void             mergeDetections(Yolo::detection *dets, int ndets, int classes, double nms_thresh=0.45, nmsKind_t nsm_kind=GREEDY_NMS);
 };

 /**
@@ -65,7 +65,7 @@ public:
    MobilenetDetection() {};
    ~MobilenetDetection() {}; 

-    bool init(const std::string& tensor_path, const int n_classes, const int n_batches=1);
+    bool init(const std::string& tensor_path, const int n_classes, const int n_batches=1, const float conf_thresh=0.3);
    void preprocess(cv::Mat &frame, const int bi=0);
    void postprocess(const int bi=0,const bool mAP=false);
 };
@@ -7,12 +7,12 @@
 namespace tk { namespace dnn {

 /**
-    Data rapresentation beetween layers
+    Data representation between layers
    n = batch size
    c = channels
-    h = heigth (lines)
+    h = height (lines)
    w = width  (rows)
-    l = lenght (3rd dimension)
+    l = length (3rd dimension)
 */
 struct dataDim_t {

@@ -43,7 +43,7 @@ public:
    void releaseLayers();

    /**
-        Do inferece for every added layer
+        Do inference for every added layer
    */
    dnnType* infer(dataDim_t &dim, dnnType* data);

@@ -6,6 +6,7 @@
 #include "Network.h"
 #include "Layer.h"
 #include "NvInfer.h"
+#include <memory>

 namespace tk { namespace dnn {

@@ -24,11 +25,12 @@ template<typename T> T readBUF(const char*& buffer)

 using namespace nvinfer1;
 #include "pluginsRT/ActivationLeakyRT.h"
+#include "pluginsRT/ActivationLogisticRT.h"
 #include "pluginsRT/ActivationReLUCeilingRT.h"
 #include "pluginsRT/ActivationMishRT.h"
 #include "pluginsRT/ReorgRT.h"
 #include "pluginsRT/RegionRT.h"
-//#include "pluginsRT/RouteRT.h"
+#include "pluginsRT/RouteRT.h"
 #include "pluginsRT/ShortcutRT.h"
 #include "pluginsRT/YoloRT.h"
 #include "pluginsRT/UpsampleRT.h"
@@ -59,6 +61,7 @@ public:
 #if NV_TENSORRT_MAJOR >= 6  
    nvinfer1::IBuilderConfig *configRT;
 #endif
+    
    nvinfer1::ICudaEngine *engineRT;
    nvinfer1::IExecutionContext *contextRT;

@@ -91,7 +94,7 @@ public:
    }

    /**
-        Do inferece
+        Do inference
    */
    dnnType* infer(dataDim_t &dim, dnnType* data);
    void enqueue(int batchSize = 1);    
@@ -115,6 +118,9 @@ public:

    bool serialize(const char *filename);
    bool deserialize(const char *filename);
+
+
+
 };

 }}
@@ -24,7 +24,7 @@ public:
    Yolo3Detection() {};
    ~Yolo3Detection() {}; 

-    bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1);
+    bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1, const float conf_thresh=0.3);
    void preprocess(cv::Mat &frame, const int bi=0);
    void postprocess(const int bi=0,const bool mAP=false);
 };
@@ -73,12 +73,12 @@ double computeMap(  std::vector<Frame> &images,const int classes,
 *                  all the recall levels are evaluated, otherwise only 
 *                  map_point recall levels are used. For COCO evaluation 
 *                  101 points are used.
- * @param map_step step used to increment IoU theshold
+ * @param map_step step used to increment IoU threshold
 * @param map_levels number of IoU step to perform
 * @param verbose is set to true, prints on screen additional info
 * @param write_on_file if set to true, the results produced by this function 
 *                      are written on file
- * @param net name of the considerd neural network 
+ * @param net name of the considered neural network 
 *
 * @return  mAP IoU_tresh:IoU_tresh+map_step*map_levels (e.g. mAP 0.5:0.95 when 
 *          map_step=0.05 and map_levels=10)
@@ -89,7 +89,7 @@ double computeMapNIoULevels(std::vector<Frame> &images,const int classes,
                const int map_levels=10, const bool verbose=false, 
                const bool write_on_file = false, std::string net = "");
 /**
- * This method computes the numper of True Positive (TP), False Positive (FP),
+ * This method computes the number of True Positive (TP), False Positive (FP),
 * False Negative (FN), precision, recall and f1-score.
 * Those values are computer over all the detections, over all the classes.
 *
@@ -101,7 +101,7 @@ double computeMapNIoULevels(std::vector<Frame> &images,const int classes,
 * @param verbose is set to true, prints on screen additional info
 * @param write_on_file if set to true, the results produced by this function 
 *                      are written on file
- * @param net name of the considerd neural network 
+ * @param net name of the considered neural network 
 */
 void computeTPFPFN( std::vector<Frame> &images,const int classes,
                    const float IoU_thresh=0.5, const float conf_thresh=0.3, 
@@ -51,9 +51,9 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
-		tk::dnn::writeBUF(buf, slope);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, size);
+		assert(buf == a + getSerializationSize());
 	}

 	int size;
@@ -0,0 +1,60 @@
+#include<cassert>
+#include "../kernels.h"
+
+class ActivationLogisticRT : public IPlugin {
+
+public:
+    ActivationLogisticRT() {
+
+
+    }
+
+    ~ActivationLogisticRT(){
+
+    }
+
+    int getNbOutputs() const override {
+        return 1;
+    }
+
+    Dims getOutputDimensions(int index, const Dims* inputs, int nbInputDims) override {
+        return inputs[0];
+    }
+
+    void configure(const Dims* inputDims, int nbInputs, const Dims* outputDims, int nbOutputs, int maxBatchSize) override {
+        size = 1;
+        for(int i=0; i<outputDims[0].nbDims; i++)
+            size *= outputDims[0].d[i];
+    }
+
+    int initialize() override {
+
+        return 0;
+    }
+
+    virtual void terminate() override {
+    }
+
+    virtual size_t getWorkspaceSize(int maxBatchSize) const override {
+        return 0;
+    }
+
+    virtual int enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream) override {
+
+        activationLOGISTICForward((dnnType*)reinterpret_cast<const dnnType*>(inputs[0]),
+                                  reinterpret_cast<dnnType*>(outputs[0]), batchSize*size, stream);
+        return 0;
+    }
+
+
+    virtual size_t getSerializationSize() override {
+        return 1*sizeof(int);
+    }
+
+    virtual void serialize(void* buffer) override {
+        char *buf = reinterpret_cast<char*>(buffer);
+        tk::dnn::writeBUF(buf, size);
+    }
+
+    int size;
+};
@@ -52,8 +52,9 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, size);
+		assert(buf == a + getSerializationSize());
 	}

 	int size;
@@ -51,9 +51,10 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, ceiling);
 		tk::dnn::writeBUF(buf, size);
+		assert(buf = a + getSerializationSize());
 		
 	}

@@ -52,8 +52,9 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, size);
+		assert(buf == a + getSerializationSize());
 	}

 	int size;
@@ -89,7 +89,7 @@ public:
 		for(int b=0; b<batchSize; b++) {
 			checkCuda(cudaMemcpy(offset, output_conv + b * 3 * chunk_dim, 2*chunk_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice)); 
 			checkCuda(cudaMemcpy(mask, output_conv + b * 3 * chunk_dim + 2*chunk_dim, chunk_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice)); 
-			// kernel sigmoide
+			// kernel sigmoid
 			activationSIGMOIDForward(mask, mask, chunk_dim);
 			// deformable convolution
 			dcnV2CudaForward(stat, handle, 
@@ -116,7 +116,7 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, chunk_dim);
 		tk::dnn::writeBUF(buf, kh);
 		tk::dnn::writeBUF(buf, kw);
@@ -163,6 +163,7 @@ public:
        for(int i=0; i<dim_ones; i++)
    		tk::dnn::writeBUF(buf, aus[i]);
 		free(aus);
+		assert(buf == a + getSerializationSize());
 	}

 	cublasStatus_t stat; 
@@ -65,12 +65,13 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a = buf;
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
 		tk::dnn::writeBUF(buf, rows);
 		tk::dnn::writeBUF(buf, cols);
+		assert(buf == a + getSerializationSize());
 	}

 	int c, h, w;
@@ -55,7 +55,7 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;

 		tk::dnn::writeBUF(buf, this->c);
 		tk::dnn::writeBUF(buf, this->h);
@@ -65,6 +65,7 @@ public:
 		tk::dnn::writeBUF(buf, this->stride_W);
 		tk::dnn::writeBUF(buf, this->winSize);
 		tk::dnn::writeBUF(buf, this->padding);
+		assert(buf == a + getSerializationSize());
 	}

 	int n, c, h, w;
@@ -73,13 +73,14 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, classes);
 		tk::dnn::writeBUF(buf, coords);
 		tk::dnn::writeBUF(buf, num);
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
+		assert(buf == a + getSerializationSize());
 	}

 	int c, h, w;
@@ -52,11 +52,12 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, stride);
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
+		assert(buf == a + getSerializationSize());
 	}

 	int c, h, w, stride;
@@ -50,11 +50,12 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a = buf;
 		tk::dnn::writeBUF(buf, n);
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
+		assert(buf == a + getSerializationSize());
 	}

 	int n, c, h, w;
@@ -52,7 +52,7 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;

 		tk::dnn::writeBUF(buf, o_c);
 		tk::dnn::writeBUF(buf, o_h);
@@ -61,6 +61,7 @@ public:
 		tk::dnn::writeBUF(buf, i_c);
 		tk::dnn::writeBUF(buf, i_h);
 		tk::dnn::writeBUF(buf, i_w);
+		assert(buf == a + getSerializationSize());
 	}

 	int i_c, i_h, i_w, o_c, o_h, o_w;
@@ -8,7 +8,9 @@ class RouteRT : public IPlugin {
 	*/

 public:
-	RouteRT() {
+	RouteRT(int groups, int group_id) {
+		this->groups = groups;
+		this->group_id = group_id;
 	}

 	~RouteRT(){
@@ -22,7 +24,7 @@ public:
 	Dims getOutputDimensions(int index, const Dims* inputs, int nbInputDims) override {
 		int out_c = 0;
 		for(int i=0; i<nbInputDims; i++) out_c += inputs[i].d[0];
-		return DimsCHW{out_c, inputs[0].d[1], inputs[0].d[2]};
+		return DimsCHW{out_c/groups, inputs[0].d[1], inputs[0].d[2]};
 	}

 	void configure(const Dims* inputDims, int nbInputs, const Dims* outputDims, int nbOutputs, int maxBatchSize) override {
@@ -34,6 +36,7 @@ public:
 		}
 		h = inputDims[0].d[1];
 		w = inputDims[0].d[2];
+		c /= groups;
 	}

 	int initialize() override {
@@ -49,15 +52,18 @@ public:
 	}

 	virtual int enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream) override {
-
+		
 		dnnType *dstData = reinterpret_cast<dnnType*>(outputs[0]);

-		int offset = 0;
-		for(int i=0; i<in; i++) {
-			dnnType *input = (dnnType*)reinterpret_cast<const dnnType*>(inputs[i]);
-			int in_dim = c_in[i]*h*w;
-			checkCuda( cudaMemcpyAsync(dstData + offset, input, in_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice, stream) );
-			offset += in_dim;
+		for(int b=0; b<batchSize; b++) {
+			int offset = 0;
+			for(int i=0; i<in; i++) {
+				dnnType *input = (dnnType*)reinterpret_cast<const dnnType*>(inputs[i]);
+				int in_dim = c_in[i]*h*w;
+				int part_in_dim = in_dim / this->groups;
+				checkCuda( cudaMemcpyAsync(dstData + b*c*w*h + offset, input + b*c*w*h*groups + this->group_id*part_in_dim, part_in_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice, stream) );
+				offset += part_in_dim;
+			}
 		}

 		return 0;
@@ -65,11 +71,13 @@ public:


 	virtual size_t getSerializationSize() override {
-		return (4+MAX_INPUTS)*sizeof(int);
+		return (6+MAX_INPUTS)*sizeof(int);
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
+		tk::dnn::writeBUF(buf, groups);
+		tk::dnn::writeBUF(buf, group_id);
 		tk::dnn::writeBUF(buf, in);
 		for(int i=0; i<MAX_INPUTS; i++)
 			tk::dnn::writeBUF(buf, c_in[i]);
@@ -77,10 +85,12 @@ public:
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
+		assert(buf == a + getSerializationSize());
 	}

 	static const int MAX_INPUTS = 4;
 	int in;
 	int c_in[MAX_INPUTS];
 	int c, h, w;
+	int groups, group_id;
 };
@@ -59,7 +59,7 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, bc);
 		tk::dnn::writeBUF(buf, bh);
 		tk::dnn::writeBUF(buf, bw);
@@ -67,7 +67,8 @@ public:
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
-			
+		assert(buf == a + getSerializationSize());
+		
 	}

 	int c, h, w;
@@ -54,11 +54,12 @@ public:
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
 		tk::dnn::writeBUF(buf, stride);
 		tk::dnn::writeBUF(buf, c);
 		tk::dnn::writeBUF(buf, h);
 		tk::dnn::writeBUF(buf, w);
+		assert(buf == a + getSerializationSize());
 	}

 	int c, h, w, stride;
@@ -8,12 +8,15 @@ class YoloRT : public IPlugin {


 public:
-	YoloRT(int classes, int num, tk::dnn::Yolo *yolo = nullptr, int n_masks=3, float scale_xy=1) {
+	YoloRT(int classes, int num, tk::dnn::Yolo *yolo = nullptr, int n_masks=3, float scale_xy=1, float nms_thresh=0.45, int nms_kind=0, int new_coords=0) {

 		this->classes = classes;
 		this->num = num;
 		this->n_masks = n_masks;
 		this->scaleXY = scale_xy;
+		this->nms_thresh = nms_thresh;
+		this->nms_kind = nms_kind;
+		this->new_coords = new_coords;

        mask = new dnnType[n_masks];
        bias = new dnnType[num*n_masks*2];
@@ -61,17 +64,23 @@ public:

 		checkCuda( cudaMemcpyAsync(dstData, srcData, batchSize*c*h*w*sizeof(dnnType), cudaMemcpyDeviceToDevice, stream));

-		for (int b = 0; b < batchSize; ++b){
-			for(int n = 0; n < n_masks; ++n){
-				int index = entry_index(b, n*w*h, 0);
-				activationLOGISTICForward(srcData + index, dstData + index, 2*w*h, stream);

-				if (this->scaleXY != 1) scalAdd(dstData + index, 2 * w*h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
-				
-				index = entry_index(b, n*w*h, 4);
-				activationLOGISTICForward(srcData + index, dstData + index, (1+classes)*w*h, stream);
-			}
-		}
+        for (int b = 0; b < batchSize; ++b){
+            for(int n = 0; n < n_masks; ++n){
+                int index = entry_index(b, n*w*h, 0);
+                if (new_coords == 1){
+                    if (this->scaleXY != 1) scalAdd(dstData + index, 2 * w*h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
+                }
+                else{
+                    activationLOGISTICForward(srcData + index, dstData + index, 2*w*h, stream); //x,y
+
+                    if (this->scaleXY != 1) scalAdd(dstData + index, 2 * w*h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
+
+                    index = entry_index(b, n*w*h, 4);
+                    activationLOGISTICForward(srcData + index, dstData + index, (1+classes)*w*h, stream);
+                }
+            }
+        }

 		//std::cout<<"YOLO END\n";
 		return 0;
@@ -79,22 +88,29 @@ public:


 	virtual size_t getSerializationSize() override {
-		return 6*sizeof(int) + sizeof(float)+ n_masks*sizeof(dnnType) + num*n_masks*2*sizeof(dnnType) + YOLORT_CLASSNAME_W*classes*sizeof(char);
+		return 8*sizeof(int) + 2*sizeof(float)+ n_masks*sizeof(dnnType) + num*n_masks*2*sizeof(dnnType) + YOLORT_CLASSNAME_W*classes*sizeof(char);
 	}

 	virtual void serialize(void* buffer) override {
-		char *buf = reinterpret_cast<char*>(buffer);
-		tk::dnn::writeBUF(buf, classes);
-		tk::dnn::writeBUF(buf, num);
-		tk::dnn::writeBUF(buf, n_masks);
-		tk::dnn::writeBUF(buf, c);
-		tk::dnn::writeBUF(buf, h);
-		tk::dnn::writeBUF(buf, w);
-		tk::dnn::writeBUF(buf, scaleXY);
-        for(int i=0; i<n_masks; i++)
-    		tk::dnn::writeBUF(buf, mask[i]);
-        for(int i=0; i<n_masks*2*num; i++)
-    		tk::dnn::writeBUF(buf, bias[i]);
+		char *buf = reinterpret_cast<char*>(buffer),*a=buf;
+		tk::dnn::writeBUF(buf, classes); 	//std::cout << "Classes :" << classes << std::endl;
+		tk::dnn::writeBUF(buf, num); 		//std::cout << "Num : " << num << std::endl;
+		tk::dnn::writeBUF(buf, n_masks); 	//std::cout << "N_Masks" << n_masks << std::endl;
+		tk::dnn::writeBUF(buf, scaleXY); 	//std::cout << "ScaleXY :" << scaleXY << std::endl;
+		tk::dnn::writeBUF(buf, nms_thresh); //std::cout << "nms_thresh :" << nms_thresh << std::endl;
+		tk::dnn::writeBUF(buf, nms_kind); 	//std::cout << "nms_kind : " << nms_kind << std::endl;
+		tk::dnn::writeBUF(buf, new_coords); //std::cout << "new_coords : " << new_coords << std::endl;
+		tk::dnn::writeBUF(buf, c); 			//std::cout << "C : " << c << std::endl;
+		tk::dnn::writeBUF(buf, h); 			//std::cout << "H : " << h << std::endl;
+		tk::dnn::writeBUF(buf, w); 			//std::cout << "C : " << c << std::endl;
+		for (int i = 0; i < n_masks; i++)
+		{
+			tk::dnn::writeBUF(buf, mask[i]); //std::cout << "mask[i] : " << mask[i] << std::endl;
+		}
+		for (int i = 0; i < n_masks * 2 * num; i++)
+		{
+			tk::dnn::writeBUF(buf, bias[i]); //std::cout << "bias[i] : " << bias[i] << std::endl;
+		}

 		// save classes names
 		for(int i=0; i<classes; i++) {
@@ -104,11 +120,15 @@ public:
 				tk::dnn::writeBUF(buf, tmp[j]);
 			}
 		}
+		assert(buf == a + getSerializationSize());
 	}

 	int c, h, w;
    int classes, num, n_masks;
 	float scaleXY;
+	float nms_thresh;
+	int nms_kind;
+	int new_coords;
 	std::vector<std::string> classesNames;

    dnnType *mask;
@@ -20,7 +20,7 @@ int testInference(std::vector<std::string> input_bins, std::vector<std::string>
    }
    if(output_bins.size() != outputs.size()) {
        std::cout<<output_bins.size()<<" "<<outputs.size()<<"\n";
-        FatalError("outputs size missmatch");
+        FatalError("outputs size mismatch");
    }

    // Load input
@@ -29,7 +29,8 @@ int testInference(std::vector<std::string> input_bins, std::vector<std::string>
    readBinaryFile(input_bins[0], net->input_dim.tot(), &input_h, &data);

    // outputs
-    dnnType *cudnn_out[outputs.size()], *rt_out[outputs.size()]; 
+    //dnnType *cudnn_out[outputs.size()], *rt_out[outputs.size()];
+    std::vector<dnnType *> cudnn_out,rt_out;

    tk::dnn::dataDim_t dim1 =  net->input_dim; //input dim
    printCenteredTitle(" CUDNN inference ", '=', 30); {
@@ -39,7 +40,7 @@ int testInference(std::vector<std::string> input_bins, std::vector<std::string>
        TKDNN_TSTOP
        dim1.print();   
    }
-    for(int i=0; i<outputs.size(); i++) cudnn_out[i] = outputs[i]->dstData;
+    for(int i=0; i<outputs.size(); i++) cudnn_out.push_back(outputs[i]->dstData);

    if(netRT != nullptr) {
        tk::dnn::dataDim_t dim2 = net->input_dim;
@@ -50,7 +51,7 @@ int testInference(std::vector<std::string> input_bins, std::vector<std::string>
            TKDNN_TSTOP
            dim2.print();
        }
-        for(int i=0; i<outputs.size(); i++) rt_out[i] = (dnnType*)netRT->buffersRT[i+1];
+        for(int i=0; i<outputs.size(); i++) rt_out.push_back((dnnType*)netRT->buffersRT[i+1]);
    }

    int ret_cudnn = 0, ret_tensorrt = 0, ret_cudnn_tensorrt = 0; 
@@ -12,8 +12,12 @@
 #include <cublas_v2.h>
 #include <cudnn.h>

+#ifdef __linux__
 #include <unistd.h>
+#endif
+
 #include <ios>
+#include <chrono>


 #define dnnType float
@@ -39,6 +43,7 @@
 #define TKDNN_VERBOSE 1

 // Simple Timer 
+#ifdef __linux__
 #define TKDNN_TSTART timespec start, end;                               \
                    clock_gettime(CLOCK_MONOTONIC, &start);            

@@ -48,6 +53,14 @@
    if(show) std::cout<<col<<"Time:"<<std::setw(16)<<t_ns<<" ms\n"<<COL_END; 

 #define TKDNN_TSTOP TKDNN_TSTOP_C(COL_CYANB, TKDNN_VERBOSE)
+#elif _WIN32
+#define TKDNN_TSTART auto start = std::chrono::high_resolution_clock::now();
+#define TKDNN_TSTOP auto stop = std::chrono::high_resolution_clock::now(); \
+std::chrono::duration<double> duration = stop -start;                      \
+auto time_ms = std::chrono::duration_cast<std::chrono::milliseconds>(duration);\
+double t_ns = time_ms.count();
+#endif
+

 /********************************************************
 * Prints the error message, and exits
@@ -0,0 +1,39 @@
+import os
+import urllib.request as dowReq
+import zipfile
+
+val = input("Enter BDD or COCO :")
+if(val == "COCO"):
+    url = "https://cloud.hipert.unimore.it/s/LNxBDk4wzqXPL8c/download"
+    lib = "..\demo\COCO_val2017"
+    lib_zip = "COCO_val2017.zip"
+elif(val == "BDD"):
+    url = "https://cloud.hipert.unimore.it/s/bikqk3FzCq2tg4D/download"
+    lib = "..\demo\BDD100k_val"
+    lib_zip = "BDD100k_val.zip"
+
+dowReq.urlretrieve(url,lib_zip)
+
+with zipfile.ZipFile(lib_zip,'r') as zip_ref:
+    zip_ref.extractall(lib)
+
+labelFolder = lib + "\labels"
+imageFolder = lib + "\images"
+
+file1 = open(".\\..\\demo\\all_labels.txt","a")
+path1 = os.path.realpath(labelFolder)
+for file in os.listdir(labelFolder):
+    valTemp = path1 + "\\" + file
+    valTemp = valTemp + '\n'
+    file1.write(valTemp)
+file1.close()
+
+file2 = open(".\\..\\demo\\all_images.txt","a")
+path2 = os.path.realpath(imageFolder)
+for file in os.listdir(imageFolder):
+    pathtemp = path2 + "\\" + file
+    pathtemp = pathtemp + '\n'
+    file2.write(pathtemp)
+file2.close()
+
+print("Completed")
@@ -69,12 +69,16 @@ do
    echo -e "${ORANGE}Batch $TKDNN_BATCHSIZE ${NC}"
    
    test_net mnist    
-    ./test_imuodom &>> $out_file
-    print_output $? imuodom
+    # ./test_imuodom &>> $out_file
+    # print_output $? imuodom

    test_net shelfnet
+    test_net shelfnet_berkeley
    test_net yolo4
+    test_net yolo4-csp
+    test_net yolo4x
    test_net yolo4_berkeley
+    test_net yolo4tiny
    test_net yolo3
    test_net yolo3_berkeley
    test_net yolo3_coco4
@@ -52,6 +52,10 @@ dnnType* Activation::infer(dataDim_t &dim, dnnType* srcData) {
    else if(act_mode == ACTIVATION_MISH) {
        activationMishForward(srcData, dstData, dim.tot());

+    }
+    else if(act_mode == ACTIVATION_LOGISTIC) {
+        activationLOGISTICForward(srcData, dstData, dim.tot());
+
    } else {
        dnnType alpha = dnnType(1);
        dnnType beta  = dnnType(0);
@@ -3,11 +3,12 @@

 namespace tk { namespace dnn {

-bool CenternetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches){
+bool CenternetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches, const float conf_thresh){
    std::cout<<(tensor_path).c_str()<<"\n";
    netRT = new tk::dnn::NetworkRT(NULL, (tensor_path).c_str() );
    classes = n_classes;
    nBatches = n_batches;
+    confThreshold = conf_thresh;

    dim = netRT->input_dim;

@@ -371,10 +372,10 @@ void CenternetDetection::postprocess(const int bi, const bool mAP){
                    // std::cout<<"th: "<<scores[j]<<" - cl: "<<clses[j]<<" i: "<<i<<std::endl;
                    //add coco bbox
                    //det[0:4], i, det[4]
-                    int x0   = target_coords[j*4];
-                    int y0   = target_coords[j*4+1];
-                    int x1   = target_coords[j*4+2];
-                    int y1   = target_coords[j*4+3];
+                    float x0   = target_coords[j*4];
+                    float y0   = target_coords[j*4+1];
+                    float x1   = target_coords[j*4+2];
+                    float y1   = target_coords[j*4+3];
                    int obj_class = clses[j];
                    float prob = scores[j];
                    // std::cout<<"("<<x0<<", "<<y0<<"),("<<x1<<", "<<y1<<")"<<std::endl;
@@ -62,25 +62,30 @@ void Conv2d::initCUDNN(bool back) {
    // init workspace
    workSpace = NULL;
    ws_sizeInBytes = 0;
+    int algo_count = 0;
    if(back) {
-        checkCUDNN( cudnnGetConvolutionBackwardDataAlgorithm(net->cudnnHandle,
-                                                        filterDesc, dstTensor, convDesc, srcTensor,
-                                                        CUDNN_CONVOLUTION_BWD_DATA_PREFER_FASTEST, 0, &bwAlgo) );
+        checkCUDNN( cudnnGetConvolutionBackwardDataAlgorithm_v7(net->cudnnHandle,
+                                                                filterDesc, dstTensor, convDesc, srcTensor, 1, &algo_count, &bwAlgo) );
        checkCUDNN(cudnnGetConvolutionBackwardDataWorkspaceSize(net->cudnnHandle,
-                                                           filterDesc, dstTensor, convDesc, srcTensor,
-                                                           bwAlgo, &ws_sizeInBytes));
+                                                                filterDesc, dstTensor, convDesc, srcTensor,
+                                                                bwAlgo.algo, &ws_sizeInBytes));
+        

        // invert tensors
        srcTensorDesc = dstTensor;
        dstTensorDesc = srcTensor;
    } else {
-        checkCUDNN( cudnnGetConvolutionForwardAlgorithm(net->cudnnHandle,
-                                                        srcTensor, filterDesc, convDesc, dstTensor,
-                                                        CUDNN_CONVOLUTION_FWD_PREFER_FASTEST, 0, &algo) );
-        checkCUDNN(cudnnGetConvolutionForwardWorkspaceSize(net->cudnnHandle,
-                                                           srcTensor, filterDesc, convDesc, dstTensor,
-                                                           algo, &ws_sizeInBytes));
+        
+        checkCUDNN( cudnnGetConvolutionForwardAlgorithm_v7(net->cudnnHandle,
+                                                         srcTensor, filterDesc, convDesc, dstTensor,
+                                                        1, &algo_count, &algo) );
+         checkCUDNN(cudnnGetConvolutionForwardWorkspaceSize(net->cudnnHandle,
+                                                            srcTensor, filterDesc, convDesc, dstTensor,
+                                                           algo.algo, &ws_sizeInBytes));
    }
+
+    if(algo_count < 1)
+        FatalError("Cannot retrieve convolutional algo");
 }

 void Conv2d::inferCUDNN(dnnType* srcData, bool back) {
@@ -91,12 +96,12 @@ void Conv2d::inferCUDNN(dnnType* srcData, bool back) {
        checkCUDNN(cudnnConvolutionBackwardData(net->cudnnHandle,
                                                &alpha, filterDesc, data_d,
                                                srcTensorDesc, srcData,
-                                                convDesc, bwAlgo, workSpace, ws_sizeInBytes,
+                                                convDesc, bwAlgo.algo, workSpace, ws_sizeInBytes,
                                                &beta, dstTensorDesc, dstData));
    } else {
        checkCUDNN(cudnnConvolutionForward(net->cudnnHandle,
                                           &alpha, srcTensorDesc, srcData, filterDesc,
-                                           data_d, convDesc, algo, workSpace, ws_sizeInBytes,
+                                           data_d, convDesc, algo.algo, workSpace, ws_sizeInBytes,
                                           &beta, dstTensorDesc, dstData));
    }

@@ -37,7 +37,10 @@ namespace tk { namespace dnn {
        std::string name,value;
        if(!divideNameAndValue(line, name, value))
            return false;
-        if(name.find("width") !=  std::string::npos)
+
+        if(name.find("new_coords") !=  std::string::npos)
+            fields.new_coords = std::stoi(value);
+        else if(name.find("width") !=  std::string::npos)
            fields.width = std::stoi(value);
        else if(name.find("height") !=  std::string::npos)
            fields.height = std::stoi(value);
@@ -75,8 +78,17 @@ namespace tk { namespace dnn {
            fields.coords = std::stoi(value);
        else if(name.find("groups") !=  std::string::npos)
            fields.groups = std::stoi(value);
+        else if(name.find("group_id") !=  std::string::npos)
+            fields.group_id = std::stoi(value);
        else if(name.find("scale_x_y") !=  std::string::npos)
            fields.scale_xy = std::stof(value);
+        else if(name.find("beta_nms") !=  std::string::npos)
+            fields.nms_thresh = std::stof(value);
+        else if(name.find("nms_kind") !=  std::string::npos){
+            if(value == "greedynms") fields.nms_kind = 0;
+            else if(value == "diounms") fields.nms_kind = 1;
+            else std::cout<<"Not supported nms_kind "<<value<<", setting to greedynms"<<std::endl;
+        }
        else if(name.find("from") !=  std::string::npos)
            fields.layers.push_back(std::stof(value));
        else if(name.find("mask") !=  std::string::npos){
@@ -148,7 +160,7 @@ namespace tk { namespace dnn {
                //std::cout<<"Route to "<<layerIdx<<" "<<netLayers[layerIdx]->getLayerName()<<"\n";
                layers.push_back(netLayers[layerIdx]);
            }
-            netLayers.push_back(new tk::dnn::Route(net, layers.data(), layers.size()));
+            netLayers.push_back(new tk::dnn::Route(net, layers.data(), layers.size(), f.groups, f.group_id));

        } else if(f.type == "reorg") {
            netLayers.push_back(new tk::dnn::Reorg(net, f.stride_x));
@@ -159,7 +171,7 @@ namespace tk { namespace dnn {
        } else if(f.type == "yolo") {
            std::string wgs = wgs_path + "/g" + std::to_string(netLayers.size()) + ".bin";
            //printf("%d %d %s %d %f\n", f.classes, f.num/f.n_mask, wgs.c_str(), f.n_mask, f.scale_xy);
-            tk::dnn::Yolo *l = new tk::dnn::Yolo(net, f.classes, f.num/f.n_mask, wgs, f.n_mask, f.scale_xy);
+            tk::dnn::Yolo *l = new tk::dnn::Yolo(net, f.classes, f.num/f.n_mask, wgs, f.n_mask, f.scale_xy, f.nms_thresh, (tk::dnn::Yolo::nmsKind_t) f.nms_kind, f.new_coords);
            if(names.size() != f.classes)
                FatalError("Mismatch between number of classes and names");
            l->classesNames = names;
@@ -175,6 +187,7 @@ namespace tk { namespace dnn {
            if(f.activation == "relu") act = tkdnnActivationMode_t(CUDNN_ACTIVATION_RELU);
            else if(f.activation == "leaky") act = tk::dnn::ACTIVATION_LEAKY;
            else if(f.activation == "mish") act = tk::dnn::ACTIVATION_MISH;
+            else if(f.activation == "logistic") act = tk::dnn::ACTIVATION_LOGISTIC;
            else { FatalError("activation not supported: " + f.activation); }
            netLayers[netLayers.size()-1] = new tk::dnn::Activation(net, act);
        };
@@ -199,7 +212,7 @@ namespace tk { namespace dnn {

        tk::dnn::Network *net = nullptr;
        
-        // layers without activations to retrive correct id number
+        // layers without activations to retrieve correct id number
        std::vector<tk::dnn::Layer*> netLayers;

        std::ifstream if_cfg(cfg_file);
@@ -95,7 +95,7 @@ dnnType* DeformConv2d::infer(dataDim_t &dim, dnnType* srcData) {
    // split conv2d outputs into offset and mask
    checkCuda(cudaMemcpy(offset, output_conv, 2*chunk_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice)); 
    checkCuda(cudaMemcpy(mask, output_conv + 2*chunk_dim, chunk_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice)); 
-    // kernel sigmoide
+    // kernel sigmoid
    activationSIGMOIDForward(mask, mask, chunk_dim);

    // deformable convolution
@@ -37,7 +37,7 @@ dnnType* Dense::infer(dataDim_t &dim, dnnType* srcData) {
    // place bias into dstData
    checkCuda( cudaMemcpy(dstData, bias_d, dim_y*sizeof(dnnType), cudaMemcpyDeviceToDevice) );
    
-    //do matrix moltiplication
+    //do matrix multiplication
    checkERROR( cublasSgemv(net->cublasHandle, CUBLAS_OP_T,
                            dim_x, dim_y,
                            &alpha,
@@ -132,21 +132,14 @@ void BatchStream::readCVimage(std::string inputFileName, std::vector<float>& res

 void BatchStream::readLabels(std::string inputFileName, std::vector<float>& ris) {
    std::ifstream is(inputFileName.c_str());
-    //read only the first number: the image sub-portion class
-    while (true) {
+    
+    std::string line;
+    while (std::getline(is, line))
+    {
+        std::istringstream iss(line);
        float val;
-		is >> val;
-        if (!is) {
-            break;
-        }
-        // insert the first number and skip all others
+        if(!(iss >> val)) { break; } // error
        ris.push_back(val);
-        while( true ) {
-            char c;
-            is >> c;
-            if (is.peek() == '\n') //detect "\n"
-                break;
-        }   
    }
 }

@@ -87,17 +87,22 @@ LSTM::LSTM( Network *net, int hiddensize, bool returnSeq, std::string fname_weig
    checkCUDNN(cudnnCreateRNNDescriptor(&rnnDesc));

 #if CUDNN_MAJOR > 7
-    checkCUDNN(cudnnSetRNNDescriptor_v6(net->cudnnHandle,
+    checkCUDNN(cudnnSetRNNDescriptor_v6(net->cudnnHandle,rnnDesc, stateSize, numLayers, dropoutDesc,
+                                        cudnnRNNInputMode_t::CUDNN_LINEAR_INPUT,
+            //(bidirectional ? cudnnDirectionMode_t::CUDNN_BIDIRECTIONAL : cudnnDirectionMode_t::CUDNN_UNIDIRECTIONAL),
+                                        cudnnDirectionMode_t::CUDNN_UNIDIRECTIONAL,
+                                        cudnnRNNMode_t::CUDNN_LSTM,
+                                        cudnnRNNAlgo_t::CUDNN_RNN_ALGO_STANDARD,
+                                        net->dataType));
 #else
-    checkCUDNN(cudnnSetRNNDescriptor(net->cudnnHandle,
-#endif
-        rnnDesc, stateSize, numLayers, dropoutDesc,
+    checkCUDNN(cudnnSetRNNDescriptor(net->cudnnHandle,rnnDesc, stateSize, numLayers, dropoutDesc,
        cudnnRNNInputMode_t::CUDNN_LINEAR_INPUT,
        //(bidirectional ? cudnnDirectionMode_t::CUDNN_BIDIRECTIONAL : cudnnDirectionMode_t::CUDNN_UNIDIRECTIONAL),
        cudnnDirectionMode_t::CUDNN_UNIDIRECTIONAL,
        cudnnRNNMode_t::CUDNN_LSTM,
        cudnnRNNAlgo_t::CUDNN_RNN_ALGO_STANDARD,
        net->dataType));
+#endif


    // Get temp space sizes
@@ -133,7 +138,7 @@ LSTM::LSTM( Network *net, int hiddensize, bool returnSeq, std::string fname_weig
    output_dim = input_dim;
    output_dim.c = stateSize*(bidirectional ? 2 : 1);

-    // if retunseq is disabled only the last timestep is returned
+    // if retunseq is disabled only the last timestamp is returned
    if(!returnSeq) {
        output_dim.h = 1;
        output_dim.w = 1;
@@ -254,7 +259,7 @@ dnnType* LSTM::infer(dataDim_t &dim, dnnType* srcData) {
            rnnDesc,
            seqLen,                     // number of time steps (nT)
            x_desc_vec_.data(),         // input array of desc (nT*nC_in)
-            srcF,                    // input pointer
+            srcF,                       // input pointer
            hx_desc_,                   // initial hidden state desc     
            hx_ptr,                     // initial hidden state pointer 
            cx_desc_,                   // initial cell state desc      
@@ -281,7 +286,7 @@ dnnType* LSTM::infer(dataDim_t &dim, dnnType* srcData) {
            rnnDesc,
            seqLen,                     // number of time steps (nT)
            x_desc_vec_.data(),         // input array of desc (nT*nC_in)
-            srcB,                    // input pointer
+            srcB,                       // input pointer
            hx_desc_,                   // initial hidden state desc     
            hx_ptr,                     // initial hidden state pointer 
            cx_desc_,                   // initial cell state desc      
@@ -289,7 +294,7 @@ dnnType* LSTM::infer(dataDim_t &dim, dnnType* srcData) {
            w_desc_,                    // weights desc
            wb_ptr,                     // weights pointer
            y_desc_vec_.data(),         // output desc     (nT*nC_out)
-            dstB_NR,                       // output pointer
+            dstB_NR,                    // output pointer
            hy_desc_,                   // final hidden state desc        
            hy_ptr,                     // final hidden state pointer 
            cy_desc_,                   // final cell state desc          
@@ -307,7 +312,7 @@ dnnType* LSTM::infer(dataDim_t &dim, dnnType* srcData) {
            one_output_dim.c*sizeof(dnnType), cudaMemcpyDeviceToDevice));
    }    

-    // if retunseq is disabled only the last timestep is returned
+    // if retunseq is disabled only the last timestamp is returned
    if(returnSeq) {
        // forward transpose
        matrixTranspose(net->cublasHandle, dstF, dstData, 
@@ -106,7 +106,7 @@ LayerWgs::LayerWgs(Network *net, int inputs, int outputs,
        float2half(tmp_d, variance16_d, b_size);
        cudaMemcpy(variance16_h, variance16_d, b_size*sizeof(__half), cudaMemcpyDeviceToHost);

-        //conver scales
+        //convert scales
        float2half(scales_d, scales16_d, b_size);
        cudaMemcpy(scales16_h, scales16_d, b_size*sizeof(__half), cudaMemcpyDeviceToHost);
    
@@ -126,12 +126,13 @@ float MobilenetDetection::iou(const tk::dnn::box &a, const tk::dnn::box &b){
    return iou;
 }

-bool MobilenetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches){
+bool MobilenetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches, const float conf_thresh){
    std::cout<<(tensor_path).c_str()<<"\n";
    netRT = new tk::dnn::NetworkRT(NULL, (tensor_path).c_str());
    imageSize = netRT->input_dim.h;
    classes = n_classes;
    nBatches = n_batches;
+    confThreshold = conf_thresh;

    SSDSpec specs[N_SSDSPEC];

@@ -12,7 +12,7 @@ MulAdd::MulAdd(Network *net, dnnType mul, dnnType add) : Layer(net) {

    int size = input_dim.tot();

-    // create a vector with all value setted to add 
+    // create a vector with all value set to add 
    dnnType *add_vector_h = new dnnType[size];
    for(int i=0; i<size; i++)
        add_vector_h[i] = add;
@@ -140,6 +140,7 @@ NetworkRT::NetworkRT(Network *net, const char *name) {
        engineRT = builderRT->buildEngineWithConfig(*networkRT, *configRT);
 #else 
        engineRT = builderRT->buildCudaEngine(*networkRT);
+        //engineRT = std::shared_ptr<nvinfer1::ICudaEngine>(builderRT->buildCudaEngine(*networkRT));
 #endif
        if(engineRT == nullptr)
            FatalError("cloud not build cuda engine")
@@ -163,7 +164,7 @@ NetworkRT::NetworkRT(Network *net, const char *name) {
 	// note that indices are guaranteed to be less than IEngine::getNbBindings()
 	buf_input_idx = engineRT->getBindingIndex("data"); 
    buf_output_idx = engineRT->getBindingIndex("out");
-    std::cout<<"input idex = "<<buf_input_idx<<" -> output index = "<<buf_output_idx<<"\n";
+    std::cout<<"input index = "<<buf_input_idx<<" -> output index = "<<buf_output_idx<<"\n";


    Dims iDim = engineRT->getBindingDimensions(buf_input_idx);
@@ -226,7 +227,7 @@ ILayer* NetworkRT::convert_layer(ITensor *input, Layer *l) {
        return convert_layer(input, (Conv2d*) l);
    if(type == LAYER_POOLING)
        return convert_layer(input, (Pooling*) l);
-    if(type == LAYER_ACTIVATION || type == LAYER_ACTIVATION_CRELU || type == LAYER_ACTIVATION_LEAKY || type == LAYER_ACTIVATION_MISH)
+    if(type == LAYER_ACTIVATION || type == LAYER_ACTIVATION_CRELU || type == LAYER_ACTIVATION_LEAKY || type == LAYER_ACTIVATION_MISH || type == LAYER_ACTIVATION_LOGISTIC)
        return convert_layer(input, (Activation*) l);
    if(type == LAYER_SOFTMAX)
        return convert_layer(input, (Softmax*) l);
@@ -423,6 +424,12 @@ ILayer* NetworkRT::convert_layer(ITensor *input, Activation *l) {
        checkNULL(lRT);
        return lRT;
    }
+    else if(l->act_mode == ACTIVATION_LOGISTIC) {
+        IPlugin *plugin = new ActivationLogisticRT();
+        IPluginLayer *lRT = networkRT->addPlugin(&input, 1, *plugin);
+        checkNULL(lRT);
+        return lRT;
+    }
    else {
        FatalError("this Activation mode is not yet implemented");
        return NULL;
@@ -451,12 +458,15 @@ ILayer* NetworkRT::convert_layer(ITensor *input, Route *l) {
        // }
        // std::cout<<"\n";
    }
-    
-    IConcatenationLayer *lRT = networkRT->addConcatenation(tens, l->layers_n);
-    //IPlugin *plugin = new RouteRT();
-    //IPluginLayer *lRT = networkRT->addPlugin(tens, l->layers_n, *plugin);
-    checkNULL(lRT);

+    if(l->groups > 1){
+        IPlugin *plugin = new RouteRT(l->groups, l->group_id);
+        IPluginLayer *lRT = networkRT->addPlugin(tens, l->layers_n, *plugin);
+        checkNULL(lRT);
+        return lRT;
+    }
+    IConcatenationLayer *lRT = networkRT->addConcatenation(tens, l->layers_n);
+    checkNULL(lRT);
    return lRT;
 }

@@ -538,7 +548,7 @@ ILayer* NetworkRT::convert_layer(ITensor *input, Yolo *l) {
    //std::cout<<"convert Yolo\n";

    //std::cout<<"New plugin YOLO\n";
-    IPlugin *plugin = new YoloRT(l->classes, l->num, l, l->n_masks, l->scaleXY);
+    IPlugin *plugin = new YoloRT(l->classes, l->num, l, l->n_masks, l->scaleXY, l->nms_thresh, l->nsm_kind, l->new_coords);
    IPluginLayer *lRT = networkRT->addPlugin(&input, 1, *plugin);
    checkNULL(lRT);
    return lRT;
@@ -570,7 +580,7 @@ ILayer* NetworkRT::convert_layer(ITensor *input, DeformConv2d *l) {
    IPluginLayer *lRT = networkRT->addPlugin(inputs, 2, *plugin);
    checkNULL(lRT);
    lRT->setName( ("Deformable" + std::to_string(l->id)).c_str() );
-    delete(inputs);
+    delete[](inputs);
    // batchnorm
    void *bias_b, *power_b, *mean_b, *variance_b, *scales_b;
    if(dtRT == DataType::kHALF) {
@@ -647,7 +657,7 @@ bool NetworkRT::deserialize(const char *filename) {


 IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialData, size_t serialLength) {
-    const char * buf = reinterpret_cast<const char*>(serialData);
+    const char * buf = reinterpret_cast<const char*>(serialData),*bufCheck = buf;

    std::string name(layerName);
    //std::cout<<name<<std::endl;
@@ -655,35 +665,53 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
    if(name.find("ActivationLeaky") == 0) {
        ActivationLeakyRT *a = new ActivationLeakyRT(readBUF<float>(buf));
        a->size = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return a;
    }
    if(name.find("ActivationMish") == 0) {
        ActivationMishRT *a = new ActivationMishRT();
        a->size = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
+        return a;
+    }
+    if(name.find("ActivationLogistic") == 0) {
+        ActivationLogisticRT *a = new ActivationLogisticRT();
+        a->size = readBUF<int>(buf);
+        return a;
+    }
+    if(name.find("ActivationLogistic") == 0) {
+        ActivationLogisticRT *a = new ActivationLogisticRT();
+        a->size = readBUF<int>(buf);
        return a;
    }
    if(name.find("ActivationCReLU") == 0) {
-        ActivationReLUCeiling *a = new ActivationReLUCeiling(readBUF<float>(buf));
+        float activationReluTemp = readBUF<float>(buf);
+        ActivationReLUCeiling* a = new ActivationReLUCeiling(activationReluTemp);
        a->size = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return a;
    }

    if(name.find("Region") == 0) {
-        RegionRT *r = new RegionRT(readBUF<int>(buf),    //classes
-                                    readBUF<int>(buf),    //coords
-                                    readBUF<int>(buf));   //num
+        int classesTemp = readBUF<int>(buf);
+        int coordsTemp = readBUF<int>(buf);
+        int numTemp = readBUF<int>(buf);
+        RegionRT* r = new RegionRT(classesTemp, coordsTemp, numTemp);

        r->c = readBUF<int>(buf);
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 

    if(name.find("Reorg") == 0) {
-        ReorgRT *r = new ReorgRT(readBUF<int>(buf)); //stride
+        int strideTemp = readBUF<int>(buf);
+        ReorgRT *r = new ReorgRT(strideTemp);
        r->c = readBUF<int>(buf);
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 

@@ -699,27 +727,34 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
        return r;
+        assert(buf == bufCheck + serialLength);
    } 

    if(name.find("Pooling") == 0) {
-        MaxPoolFixedSizeRT *r = new MaxPoolFixedSizeRT( readBUF<int>(buf), //c
-                                                        readBUF<int>(buf), //h
-                                                        readBUF<int>(buf), //w
-                                                        readBUF<int>(buf), //n
-                                                        readBUF<int>(buf), //strideH
-                                                        readBUF<int>(buf), //strideW
-                                                        readBUF<int>(buf), //winSize
-                                                        readBUF<int>(buf)); //padding
+        int cTemp = readBUF<int>(buf);
+        int hTemp = readBUF<int>(buf);
+        int wTemp = readBUF<int>(buf);
+        int nTemp = readBUF<int>(buf);
+        int strideHTemp = readBUF<int>(buf);
+        int strideWTemp = readBUF<int>(buf);
+        int winSizeTemp = readBUF<int>(buf);
+        int paddingTemp = readBUF<int>(buf);
+
+        MaxPoolFixedSizeRT* r = new MaxPoolFixedSizeRT(cTemp, hTemp, wTemp, nTemp, strideHTemp, strideWTemp, winSizeTemp, paddingTemp);
+        assert(buf == bufCheck + serialLength);
        return r;
    }

    if(name.find("Resize") == 0) {
-        ResizeLayerRT *r = new ResizeLayerRT(readBUF<int>(buf), //o_c
-                                            readBUF<int>(buf), //o_h
-                                            readBUF<int>(buf)); //o_w
+        int o_cTemp = readBUF<int>(buf);
+        int o_hTemp = readBUF<int>(buf);
+        int o_wTemp = readBUF<int>(buf);
+        ResizeLayerRT* r = new ResizeLayerRT(o_cTemp, o_hTemp, o_wTemp);
+
        r->i_c = readBUF<int>(buf);
        r->i_h = readBUF<int>(buf);
        r->i_w = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 

@@ -730,6 +765,7 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
        r->w = readBUF<int>(buf);
        r->rows = readBUF<int>(buf);
        r->cols = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 

@@ -741,19 +777,28 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
        new_dim.h = readBUF<int>(buf);
        new_dim.w = readBUF<int>(buf);
        ReshapeRT *r = new ReshapeRT(new_dim); 
+        assert(buf == bufCheck + serialLength);
        
        return r;
    } 

    if(name.find("Yolo") == 0) {
-        YoloRT *r = new YoloRT(readBUF<int>(buf),    //classes
-                                readBUF<int>(buf),   //num
-                                nullptr,
-                                readBUF<int>(buf));   //n_masks
+
+        int classes_temp = readBUF<int>(buf);
+        int num_temp = readBUF<int>(buf);
+        int n_masks_temp = readBUF<int>(buf);
+        float scale_xy_temp = readBUF<float>(buf);
+        float nms_thresh_temp = readBUF<float>(buf);
+        int nms_kind_temp = readBUF<int>(buf);
+        int new_coords_temp = readBUF<int>(buf);
+
+       YoloRT *r = new YoloRT(classes_temp,num_temp,nullptr,n_masks_temp,scale_xy_temp,nms_thresh_temp,nms_kind_temp,new_coords_temp);  
+
+
+
        r->c = readBUF<int>(buf);
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
-        r->scaleXY = readBUF<float>(buf);
        for(int i=0; i<r->n_masks; i++)
            r->mask[i] = readBUF<dnnType>(buf);
        for(int i=0; i<r->n_masks*2*r->num; i++)
@@ -767,36 +812,54 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
 				tmp[j] = readBUF<char>(buf);
            r->classesNames[i] = std::string(tmp);
 		}
+        assert(buf == bufCheck + serialLength);

        yolos[n_yolos++] = r;
        return r;
    } 
    if(name.find("Upsample") == 0) {
-        UpsampleRT *r = new UpsampleRT(readBUF<int>(buf)); //stride
+        int strideTemp = readBUF<int>(buf);
+        UpsampleRT* r = new UpsampleRT(strideTemp);
        r->c = readBUF<int>(buf);
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 
-/*
+
    if(name.find("Route") == 0) {
-        RouteRT *r = new RouteRT();
+        int groupsTemp = readBUF<int>(buf);
+        int group_idTemp = readBUF<int>(buf);
+        RouteRT* r = new RouteRT(groupsTemp, group_idTemp);
        r->in = readBUF<int>(buf);
        for(int i=0; i<RouteRT::MAX_INPUTS; i++)
            r->c_in[i] = readBUF<int>(buf);
        r->c = readBUF<int>(buf);
        r->h = readBUF<int>(buf);
        r->w = readBUF<int>(buf);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 
-*/
+
    if(name.find("Deformable") == 0) {
-        DeformableConvRT *r = new DeformableConvRT(readBUF<int>(buf), readBUF<int>(buf), readBUF<int>(buf),
-                                                    readBUF<int>(buf), readBUF<int>(buf), readBUF<int>(buf),
-                                                    readBUF<int>(buf), readBUF<int>(buf),
-                                                    readBUF<int>(buf),readBUF<int>(buf),readBUF<int>(buf),readBUF<int>(buf),
-                                                    readBUF<int>(buf),readBUF<int>(buf),readBUF<int>(buf),readBUF<int>(buf),
-                                                    nullptr); 
+        int chuck_dimTemp = readBUF<int>(buf);
+        int khTemp = readBUF<int>(buf);
+        int kwTemp = readBUF<int>(buf);
+        int shTemp = readBUF<int>(buf);
+        int swTemp = readBUF<int>(buf);
+        int phTemp = readBUF<int>(buf);
+        int pwTemp = readBUF<int>(buf);
+        int deformableGroupTemp = readBUF<int>(buf);
+        int i_nTemp = readBUF<int>(buf);
+        int i_cTemp = readBUF<int>(buf);
+        int i_hTemp = readBUF<int>(buf);
+        int i_wTemp = readBUF<int>(buf);
+        int o_nTemp = readBUF<int>(buf);
+        int o_cTemp = readBUF<int>(buf);
+        int o_hTemp = readBUF<int>(buf);
+        int o_wTemp = readBUF<int>(buf);
+
+        DeformableConvRT* r = new DeformableConvRT(chuck_dimTemp, khTemp, kwTemp, shTemp, swTemp, phTemp, pwTemp, deformableGroupTemp, i_nTemp, i_cTemp, i_hTemp, i_wTemp, o_nTemp, o_cTemp, o_hTemp, o_wTemp, nullptr);
        dnnType *aus = new dnnType[r->chunk_dim*2];
        for(int i=0; i<r->chunk_dim*2; i++)
    		aus[i] = readBUF<dnnType>(buf);
@@ -827,6 +890,7 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
    		aus[i] = readBUF<dnnType>(buf);
 		checkCuda( cudaMemcpy(r->ones_d2, aus, sizeof(dnnType)*r->dim_ones, cudaMemcpyHostToDevice) );
        free(aus);
+        assert(buf == bufCheck + serialLength);
        return r;
    } 

@@ -63,7 +63,7 @@ dnnType* Region::infer(dataDim_t &dim, dnnType* srcData) {
 }


-/* Intepret class */
+/* Interpret class */
 RegionInterpret::RegionInterpret(dataDim_t input_dim, dataDim_t output_dim, 
    int classes, int coords, int num, float thresh, std::string fname_weights) {

@@ -5,7 +5,7 @@

 namespace tk { namespace dnn {

-Route::Route(Network *net, Layer **layers, int layers_n) : Layer(net) {
+Route::Route(Network *net, Layer **layers, int layers_n, int groups, int group_id) : Layer(net) {

    // copy input layers
    if(layers_n > MAX_LAYERS) {
@@ -15,6 +15,8 @@ Route::Route(Network *net, Layer **layers, int layers_n) : Layer(net) {
        this->layers[i] = layers[i];
    }
    this->layers_n = layers_n;
+    this->groups = groups;
+    this->group_id = group_id;
    
    //get dims
    output_dim.l = 1;
@@ -32,6 +34,7 @@ Route::Route(Network *net, Layer **layers, int layers_n) : Layer(net) {
        output_dim.c += layers[i]->output_dim.c;
    }

+    output_dim.c /= this->groups;
    input_dim = output_dim;

    checkCuda( cudaMalloc(&dstData, output_dim.tot()*sizeof(dnnType)) );
@@ -49,8 +52,9 @@ dnnType* Route::infer(dataDim_t &dim, dnnType* srcData) {
    for(int i=0; i<layers_n; i++) {
        dnnType *input = layers[i]->dstData;
        int in_dim = layers[i]->output_dim.tot();
-        checkCuda( cudaMemcpy(dstData + offset, input, in_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice));
-        offset += in_dim;
+        int part_in_dim = in_dim / this->groups;
+        checkCuda( cudaMemcpy(dstData + offset, input + this->group_id*part_in_dim, part_in_dim*sizeof(dnnType), cudaMemcpyDeviceToDevice));
+        offset += part_in_dim;
    }

    //update data dimensions    
@@ -9,9 +9,10 @@
 #include "Layer.h"
 #include "kernels.h"

+
 namespace tk { namespace dnn {

-Yolo::Yolo(Network *net, int classes, int num, std::string fname_weights, int n_masks, float scale_xy) : 
+Yolo::Yolo(Network *net, int classes, int num, std::string fname_weights, int n_masks, float scale_xy, double nms_thresh, nmsKind_t nsm_kind, int new_coords) : 
    Layer(net) {
    this->final = true;

@@ -19,6 +20,9 @@ Yolo::Yolo(Network *net, int classes, int num, std::string fname_weights, int n_
    this->num = num;
    this->n_masks = n_masks;
    this->scaleXY = scale_xy;
+    this->nms_thresh = nms_thresh;
+    this->nsm_kind = nsm_kind;
+    this->new_coords = new_coords;

    // load anchors
    if(fname_weights != "") {
@@ -59,12 +63,21 @@ int entry_index(int batch, int location, int entry,
           entry*input_dim.w*input_dim.h + loc;
 }

-Yolo::box get_yolo_box(float *x, float *biases, int n, int index, int i, int j, int lw, int lh, int w, int h, int stride) {
+Yolo::box get_yolo_box(float *x, float *biases, int n, int index, int i, int j, int lw, int lh, int w, int h, int stride, int new_coords) {
    Yolo::box b;
-    b.x = (i + x[index + 0*stride]) / lw;
-    b.y = (j + x[index + 1*stride]) / lh;
-    b.w = exp(x[index + 2*stride]) * biases[2*n]   / w;
-    b.h = exp(x[index + 3*stride]) * biases[2*n+1] / h;
+
+    if(new_coords == 0){
+        b.x = (i + x[index + 0*stride]) / lw;
+        b.y = (j + x[index + 1*stride]) / lh;
+        b.w = exp(x[index + 2*stride]) * biases[2*n]   / w;
+        b.h = exp(x[index + 3*stride]) * biases[2*n+1] / h;
+    }
+    else{
+        b.x = (i + x[index + 0 * stride] ) / lw;
+        b.y = (j + x[index + 1 * stride] ) / lh;
+        b.w = x[index + 2 * stride] * x[index + 2 * stride] * 4 * biases[2 * n] / w;
+        b.h = x[index + 3 * stride] * x[index + 3 * stride] * 4 * biases[2 * n + 1] / h;
+    }
    return b;
 }

@@ -75,12 +88,17 @@ dnnType* Yolo::infer(dataDim_t &dim, dnnType* srcData) {
    for (int b = 0; b < dim.n; ++b){
        for(int n = 0; n < n_masks; ++n){
            int index = entry_index(b, n*dim.w*dim.h, 0, classes, input_dim, output_dim);
-            activationLOGISTICForward(srcData + index, dstData + index, 2*dim.w*dim.h);
+            std::cout<<"new_coords"<<new_coords<<std::endl;
+            if (new_coords == 1){
+                if (this->scaleXY != 1) scalAdd(dstData + index, 2 * dim.w*dim.h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
+            }
+            else{
+                activationLOGISTICForward(srcData + index, dstData + index, 2*dim.w*dim.h);

-            if (this->scaleXY != 1) scalAdd(dstData + index, 2 * dim.w*dim.h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
-            
-            index = entry_index(b, n*dim.w*dim.h, 4, classes, input_dim, output_dim);
-            activationLOGISTICForward(srcData + index, dstData + index, (1+classes)*dim.w*dim.h);
+                if (this->scaleXY != 1) scalAdd(dstData + index, 2 * dim.w*dim.h, this->scaleXY, -0.5*(this->scaleXY - 1), 1);
+                index = entry_index(b, n*dim.w*dim.h, 4, classes, input_dim, output_dim);
+                activationLOGISTICForward(srcData + index, dstData + index, (1+classes)*dim.w*dim.h);
+            }
        }
    }

@@ -116,7 +134,7 @@ void correct_yolo_boxes(Yolo::detection *dets, int n, int w, int h, int netw, in
    }
 }

-int Yolo::computeDetections(Yolo::detection *dets, int &ndets, int netw, int neth, float thresh) {
+int Yolo::computeDetections(Yolo::detection *dets, int &ndets, int netw, int neth, float thresh, int new_coords) {

    if(predictions == nullptr)
        predictions = new dnnType[output_dim.tot()];
@@ -140,7 +158,7 @@ int Yolo::computeDetections(Yolo::detection *dets, int &ndets, int netw, int net
            if(objectness <= thresh) continue;
            int box_index  = entry_index(0, n*lw*lh + i, 0, classes, input_dim, output_dim);
            
-            dets[count].bbox = get_yolo_box(predictions, bias_h, mask_h[n], box_index, col, row, lw, lh, netw, neth, lw*lh);
+            dets[count].bbox = get_yolo_box(predictions, bias_h, mask_h[n], box_index, col, row, lw, lh, netw, neth, lw*lh, new_coords);
            dets[count].objectness = objectness;
            dets[count].classes = classes;
            for(j = 0; j < classes; ++j){
@@ -193,6 +211,32 @@ float yolo_box_iou(Yolo::box a, Yolo::box b)
    return yolo_box_intersection(a, b)/yolo_box_union(a, b);
 }

+void box_c(const Yolo::box a, const Yolo::box b, float& top, float& bot, float& left, float& right) {
+    top = (std::min)(a.y - a.h / 2, b.y - b.h / 2);
+    bot = (std::max)(a.y + a.h / 2, b.y + b.h / 2);
+    left = (std::min)(a.x - a.w / 2, b.x - b.w / 2);
+    right = (std::max)(a.x + a.w / 2, b.x + b.w / 2);
+}
+
+// https://github.com/Zzh-tju/DIoU-darknet
+// https://arxiv.org/abs/1911.08287
+float yolo_box_diou(const Yolo::box a, const Yolo::box b, const float nms_thresh=0.6)
+{
+    float top, bot, left, right;
+    box_c(a, b, top, bot, left, right);
+    float w = right - left;
+    float h = bot - top;
+    float c = w * w + h * h;
+    float iou = yolo_box_iou(a, b);
+    if (c == 0) 
+        return iou;
+    
+    float d = (a.x - b.x) * (a.x - b.x) + (a.y - b.y) * (a.y - b.y);
+    float u = pow(d / c, nms_thresh);
+    float diou_term = u;
+    return iou - diou_term;
+}
+
 int yolo_nms_comparator(const void *pa, const void *pb)
 {
    Yolo::detection a = *(Yolo::detection *)pa;
@@ -219,8 +263,7 @@ Yolo::detection *Yolo::allocateDetections(int nboxes, int classes) {
    return dets;
 }

-void Yolo::mergeDetections(Yolo::detection *dets, int ndets, int classes) {
-    double nms_thresh = 0.45;
+void Yolo::mergeDetections(Yolo::detection *dets, int ndets, int classes, double nms_thresh, nmsKind_t nsm_kind) {
    int total = ndets;

    int i, j, k;
@@ -246,13 +289,13 @@ void Yolo::mergeDetections(Yolo::detection *dets, int ndets, int classes) {
            box a = dets[i].bbox;
            for(j = i+1; j < total; ++j){
                box b = dets[j].bbox;
-                if (yolo_box_iou(a, b) > nms_thresh){
+                if (nsm_kind == GREEDY_NMS && yolo_box_iou(a, b) > nms_thresh)
+                    dets[j].prob[k] = 0;
+                else if (nsm_kind == DIOU_NMS && yolo_box_diou(a, b, nms_thresh) > nms_thresh)
                    dets[j].prob[k] = 0;
-                }
            }
        }
    }
-
 }

 }}
@@ -3,13 +3,14 @@

 namespace tk { namespace dnn {

-bool Yolo3Detection::init(const std::string& tensor_path, const int n_classes, const int n_batches) {
+bool Yolo3Detection::init(const std::string& tensor_path, const int n_classes, const int n_batches, const float conf_thresh) {

    //convert network to tensorRT
    std::cout<<(tensor_path).c_str()<<"\n";
    netRT = new tk::dnn::NetworkRT(NULL, (tensor_path).c_str() );

    nBatches = n_batches;
+    confThreshold = conf_thresh;
    tk::dnn::dataDim_t idim = netRT->input_dim;    
    idim.n = nBatches;

@@ -31,6 +32,9 @@ bool Yolo3Detection::init(const std::string& tensor_path, const int n_classes, c
        memcpy(yolo[i]->bias_h, yRT->bias, sizeof(dnnType)*num*nMasks*2);
        yolo[i]->input_dim = yolo[i]->output_dim = tk::dnn::dataDim_t(1, yRT->c, yRT->h, yRT->w);
        yolo[i]->classesNames = yRT->classesNames;
+        yolo[i]->nms_thresh = yRT->nms_thresh;
+        yolo[i]->nsm_kind = (tk::dnn::Yolo::nmsKind_t) yRT->nms_kind;
+        yolo[i]->new_coords = yRT->new_coords;
    }

    dets = tk::dnn::Yolo::allocateDetections(tk::dnn::Yolo::MAX_DETECTIONS, classes);
@@ -90,9 +94,10 @@ void Yolo3Detection::preprocess(cv::Mat &frame, const int bi){
 void Yolo3Detection::postprocess(const int bi, const bool mAP){

    //get yolo outputs
-    dnnType *rt_out[netRT->pluginFactory->n_yolos]; 
-    for(int i=0; i<netRT->pluginFactory->n_yolos; i++) 
-        rt_out[i] = (dnnType*)netRT->buffersRT[i+1] + netRT->buffersDIM[i+1].tot()*bi;
+    std::vector<float *> rt_out;
+    //dnnType *rt_out[netRT->pluginFactory->n_yolos];
+    for(int i=0; i<netRT->pluginFactory->n_yolos; i++)
+        rt_out.push_back((dnnType*)netRT->buffersRT[i+1] + netRT->buffersDIM[i+1].tot()*bi);

    float x_ratio =  float(originalSize[bi].width) / float(netRT->input_dim.w);
    float y_ratio =  float(originalSize[bi].height) / float(netRT->input_dim.h);
@@ -101,46 +106,47 @@ void Yolo3Detection::postprocess(const int bi, const bool mAP){
    nDets = 0;
    for(int i=0; i<netRT->pluginFactory->n_yolos; i++) {
        yolo[i]->dstData = rt_out[i];
-        yolo[i]->computeDetections(dets, nDets, netRT->input_dim.w, netRT->input_dim.h, confThreshold);
+        yolo[i]->computeDetections(dets, nDets, netRT->input_dim.w, netRT->input_dim.h, confThreshold, yolo[i]->new_coords);
    }
-    tk::dnn::Yolo::mergeDetections(dets, nDets, classes);
+    tk::dnn::Yolo::mergeDetections(dets, nDets, classes, yolo[0]->nms_thresh, yolo[0]->nsm_kind);

    // fill detected
    detected.clear();
    for(int j=0; j<nDets; j++) {
        tk::dnn::Yolo::box b = dets[j].bbox;
-        int x0   = (b.x-b.w/2.);
-        int x1   = (b.x+b.w/2.);
-        int y0   = (b.y-b.h/2.);
-        int y1   = (b.y+b.h/2.);
-        int obj_class = -1;
-        float prob = 0;
+        float x0   = (b.x-b.w/2.);
+        float x1   = (b.x+b.w/2.);
+        float y0   = (b.y-b.h/2.);
+        float y1   = (b.y+b.h/2.);
+
+        // convert to image coords
+        x0 = x_ratio*x0;
+        x1 = x_ratio*x1;
+        y0 = y_ratio*y0;
+        y1 = y_ratio*y1;
+        
        for(int c=0; c<classes; c++) {
            if(dets[j].prob[c] >= confThreshold) {
-                obj_class = c;
-                prob = dets[j].prob[c];
+                int obj_class = c;
+                float prob = dets[j].prob[c];
+
+                tk::dnn::box res;
+                res.cl = obj_class;
+                res.prob = prob;
+                res.x = x0;
+                res.y = y0;
+                res.w = x1 - x0;
+                res.h = y1 - y0;
+
+                // FIXME: this shuld be useless
+                // if(mAP)
+                //     for(int c=0; c<classes; c++) 
+                //         res.probs.push_back(dets[j].prob[c]);
+
+                detected.push_back(res);
            }
        }

-        if(obj_class >= 0) {
-            // convert to image coords
-            x0 = x_ratio*x0;
-            x1 = x_ratio*x1;
-            y0 = y_ratio*y0;
-            y1 = y_ratio*y1;
-              
-            tk::dnn::box res;
-            res.cl = obj_class;
-            res.prob = prob;
-            res.x = x0;
-            res.y = y0;
-            res.w = x1 - x0;
-            res.h = y1 - y0;
-            if(mAP)
-                for(int c=0; c<classes; c++) 
-                    res.probs.push_back(dets[j].prob[c]);
-            detected.push_back(res);
-        }
    }
    batchDetected.push_back(detected);
 }
@@ -63,7 +63,7 @@ double computeMap(  std::vector<Frame> &images,const int classes,
    
    int gt_checked = 0;

-    // for each detection comput IoU with groundtruth and match detetcion and 
+    // for each detection compute IoU with groundtruth and match detetcion and 
    // groundtruth with IoU greater than IoU_thresh
    for(auto &img:images){
        for(size_t i=0; i<img.det.size(); i++){
@@ -153,7 +153,7 @@ double computeMap(  std::vector<Frame> &images,const int classes,
        }
    }

-    //compute average precision for each class. Two methods are avaible, 
+    //compute average precision for each class. Two methods are available, 
    //based on map_points required
    double mean_average_precision = 0;
    double last_recall, last_precision, delta_recall;
@@ -287,7 +287,7 @@ void computeTPFPFN( std::vector<Frame> &images,const int classes,
        }
    }

-    //count all TP, FP, FN and compute precsion, recall and f1-score
+    //count all TP, FP, FN and compute precision, recall and f1-score
    double avg_precision = 0, avg_recall = 0, f1_score = 0;
    int TP = 0, FP = 0, FN = 0;
    for(size_t i=0; i<classes; i++){
@@ -18,7 +18,7 @@ inline int GET_BLOCKS(const int N)
 }


-__device__ float dmcn_im2col_bilinear(const float *bottom_data, const int data_width,
+__device__ __host__  float dmcn_im2col_bilinear(const float *bottom_data, const int data_width,
  const int height, const int width, float h, float w) {
 int h_low = floor(h);
 int w_low = floor(w);
@@ -23,14 +23,23 @@ bool fileExist(const char *fname) {
 void downloadWeightsifDoNotExist(const std::string& input_bin, const std::string& test_folder, const std::string& weights_url){
    if(!fileExist(input_bin.c_str())){
        std::string mkdir_cmd = "mkdir " + test_folder; 
-        std::string wget_cmd = "wget " + weights_url + " -O " + test_folder + "/weights.zip"; 
+        std::string wget_cmd = "curl " + weights_url + " --output " + test_folder + "/weights.zip";
+#ifdef __linux__
        std::string unzip_cmd = "unzip " + test_folder + "/weights.zip -d" + test_folder;
        std::string rm_cmd = "rm " + test_folder + "/weights.zip";
+
+#elif _WIN32
+
+        std::string unzip_cmd = "7z x " + test_folder + "/weights.zip -o" + test_folder;
+#endif 
        int err = 0;
        err = system(mkdir_cmd.c_str());
        err = system(wget_cmd.c_str());
        err = system(unzip_cmd.c_str());
+#ifdef __linux__
        err = system(rm_cmd.c_str());
+#endif
+
    }
 }

@@ -192,8 +201,12 @@ void getMemUsage(double& vm_usage_kb, double& resident_set_kb){
               >> O >> itrealvalue >> starttime >> vsize >> rss;

   stat_stream.close();
-
+#ifdef __linux__
   long page_size_kb = sysconf(_SC_PAGE_SIZE) / 1024; // in case x86-64 is configured to use 2MB pages
+#elif _WIN32
+ long page_size_kb = 4096/1024;
+#endif
+
   vm_usage_kb     = vsize / 1024.0;  
   resident_set_kb = rss * page_size_kb;
 }
@@ -0,0 +1,281 @@
+[net]
+# Testing
+#batch=1
+#subdivisions=1
+# Training
+batch=64
+subdivisions=1
+width=416
+height=416
+channels=3
+momentum=0.9
+decay=0.0005
+angle=0
+saturation = 1.5
+exposure = 1.5
+hue=.1
+
+learning_rate=0.00261
+burn_in=1000
+max_batches = 500200
+policy=steps
+steps=400000,450000
+scales=.1,.1
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers=-1
+groups=2
+group_id=1
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -1,-2
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -6,-1
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers=-1
+groups=2
+group_id=1
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -1,-2
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -6,-1
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers=-1
+groups=2
+group_id=1
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -1,-2
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[route]
+layers = -6,-1
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+##################################
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=255
+activation=linear
+
+
+
+[yolo]
+mask = 3,4,5
+anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
+classes=80
+num=6
+jitter=.3
+scale_x_y = 1.05
+cls_normalizer=1.0
+iou_normalizer=0.07
+iou_loss=ciou
+ignore_thresh = .7
+truth_thresh = 1
+random=0
+resize=1.5
+nms_kind=greedynms
+beta_nms=0.6
+
+[route]
+layers = -4
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[upsample]
+stride=2
+
+[route]
+layers = -1, 23
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=255
+activation=linear
+
+[yolo]
+mask = 1,2,3
+anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
+classes=80
+num=6
+jitter=.3
+scale_x_y = 1.05
+cls_normalizer=1.0
+iou_normalizer=0.07
+iou_loss=ciou
+ignore_thresh = .7
+truth_thresh = 1
+random=0
+resize=1.5
+nms_kind=greedynms
+beta_nms=0.6
@@ -17,8 +17,7 @@ int main() {
    std::string wgs_path  = bin_path + "/layers";
    std::string cfg_path  = std::string(TKDNN_PATH) + "/tests/darknet/cfg/csresnext50-panet-spp_berkeley.cfg";
    std::string name_path = std::string(TKDNN_PATH) + "/tests/darknet/names/berkeley.names";
-    // FIXME: wrong weights
-    // downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s//download");
+    downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s/q82qHAtqpoaFYo5/download");

    // parse darknet network
    tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
@@ -0,0 +1,4 @@
+blue-cone
+yellow-cone
+orange-cone
+big-orange-cone
@@ -0,0 +1,34 @@
+#include<iostream>
+#include<vector>
+#include "tkdnn.h"
+#include "test.h"
+#include "DarknetParser.h"
+
+int main() {
+    std::string bin_path  = "yolo4-csp";
+    std::vector<std::string> input_bins = { 
+        bin_path + "/layers/input.bin"
+    };
+    std::vector<std::string> output_bins = {
+        bin_path + "/debug/layer144_out.bin",
+        bin_path + "/debug/layer159_out.bin",
+        bin_path + "/debug/layer174_out.bin"
+    };
+    std::string wgs_path  = bin_path + "/layers";
+    std::string cfg_path  = std::string(TKDNN_PATH) + "/tests/darknet/cfg/yolo4-csp.cfg";
+    std::string name_path = std::string(TKDNN_PATH) + "/tests/darknet/names/coco.names";
+    downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s/AfzHE4BfTeEm2gH/download");
+
+    // parse darknet network
+    tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
+    net->print();
+
+    //convert network to tensorRT
+    tk::dnn::NetworkRT *netRT = new tk::dnn::NetworkRT(net, net->getNetworkRTName(bin_path.c_str()));
+    
+    int ret = testInference(input_bins, output_bins, net, netRT);
+    net->releaseLayers();
+    delete net;
+    delete netRT;
+    return ret;
+}
@@ -0,0 +1,34 @@
+#include<iostream>
+#include<vector>
+#include "tkdnn.h"
+#include "test.h"
+#include "DarknetParser.h"
+
+int main() {
+    std::string bin_path  = "yolo4_mmr";
+    std::vector<std::string> input_bins = { 
+        bin_path + "/layers/input.bin"
+    };
+    std::vector<std::string> output_bins = {
+        bin_path + "/debug/layer139_out.bin",
+        bin_path + "/debug/layer150_out.bin",
+        bin_path + "/debug/layer161_out.bin"
+    };
+    std::string wgs_path  = bin_path + "/layers";
+    std::string cfg_path  = std::string(TKDNN_PATH) + "/tests/darknet/cfg/yolo4_mmr.cfg";
+    std::string name_path = std::string(TKDNN_PATH) + "/tests/darknet/names/mmr.names";
+    // downloadWeightsifDoNotExist(input_bins[0], bin_path, "");
+
+    // parse darknet network
+    tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
+    net->print();
+
+    //convert network to tensorRT
+    tk::dnn::NetworkRT *netRT = new tk::dnn::NetworkRT(net, net->getNetworkRTName(bin_path.c_str()));
+    
+    int ret = testInference(input_bins, output_bins, net, netRT);
+    net->releaseLayers();
+    delete net;
+    delete netRT;
+    return ret;
+}
@@ -0,0 +1,33 @@
+#include<iostream>
+#include<vector>
+#include "tkdnn.h"
+#include "test.h"
+#include "DarknetParser.h"
+
+int main() {
+    std::string bin_path  = "yolo4tiny";
+    std::vector<std::string> input_bins = { 
+        bin_path + "/layers/input.bin"
+    };
+    std::vector<std::string> output_bins = {
+        bin_path + "/debug/layer30_out.bin",
+        bin_path + "/debug/layer37_out.bin"
+    };
+    std::string wgs_path  = bin_path + "/layers";
+    std::string cfg_path  = std::string(TKDNN_PATH) + "/tests/darknet/cfg/yolo4tiny.cfg";
+    std::string name_path = std::string(TKDNN_PATH) + "/tests/darknet/names/coco.names";
+    downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s/iRnc4pSqmx78gJs/download");
+
+    // parse darknet network
+    tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
+    net->print();
+
+    //convert network to tensorRT
+    tk::dnn::NetworkRT *netRT = new tk::dnn::NetworkRT(net, net->getNetworkRTName(bin_path.c_str()));
+    
+    int ret = testInference(input_bins, output_bins, net, netRT);
+    net->releaseLayers();
+    delete net;
+    delete netRT;
+    return ret;
+}
@@ -0,0 +1,36 @@
+#include<iostream>
+#include<vector>
+#include "tkdnn.h"
+#include "test.h"
+#include "DarknetParser.h"
+
+int main() {
+    std::string bin_path  = "yolo4x";
+    std::vector<std::string> input_bins = { 
+        bin_path + "/layers/input.bin"
+    };
+    std::vector<std::string> output_bins = {
+        bin_path + "/debug/layer168_out.bin",
+        bin_path + "/debug/layer185_out.bin",
+        bin_path + "/debug/layer202_out.bin"
+    };
+    std::string wgs_path  = bin_path + "/layers";
+    std::string cfg_path  = std::string(TKDNN_PATH) + "/tests/darknet/cfg/yolo4x.cfg";
+    std::string name_path = std::string(TKDNN_PATH) + "/tests/darknet/names/coco.names";
+    downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s/5MFjtNtgbDGdJEo/download");
+
+
+
+    // parse darknet network
+    tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
+    net->print();
+
+    //convert network to tensorRT
+    tk::dnn::NetworkRT *netRT = new tk::dnn::NetworkRT(net, net->getNetworkRTName(bin_path.c_str()));
+    
+    int ret = testInference(input_bins, output_bins, net, netRT);
+    net->releaseLayers();
+    delete net;
+    delete netRT;
+    return ret;
+}
@@ -83,7 +83,8 @@ const char *trans[] = {
 int main()
 {

-    downloadWeightsifDoNotExist(input_bin, "shelfnet_mapillary", "https://cloud.hipert.unimore.it/s/6WnZCKLjik7xrny/download");
+    // downloadWeightsifDoNotExist(input_bin, "shelfnet_mapillary", "");
+    // download the weights from here:  https://cloud.hipert.unimore.it/f/652476

    // Mapillary Vistas has originally 66 classes, but we reduced them to 15 to improve the results on the categories of our interest.
    int classes = 15;
				`@@ -0,0 +1 @@`
				`1)error C2131 @ Yolo3Detection.cpp(97) -> expression doesnt evaluate to a constant caused to read of variable outside its lifetime`