mmr/tkDNN
Archived
1
0
Fork 0
Code Issues 46 Pull Requests 12 Releases 4 Activity

Code cleanup and readme fixes

Browse Source
This commit is contained in:
perseusdg
2021-04-09 13:17:41 +05:30
parent 37b2a5bd98
commit 1de804f98d
8 changed files with 37 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CMakeLists.txt
+3 -3
View File
@@ -3,11 +3,11 @@ cmake_minimum_required(VERSION 3.5)
project (tkDNN) project (tkDNN)
set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/cmake) set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
if(UNIX) if(UNIX)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -fPIC -Wno-deprecated-declarations -Wno-unused-variable") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fPIC -Wno-deprecated-declarations -Wno-unused-variable -g ")
endif() endif()
if(WIN32) if(WIN32)
set(CMAKE_CXX_STANDARD 14) set(CMAKE_CXX_STANDARD 11)
set(CMAKE_CXX_FLAGS "/O1 /FS /EHsc") set(CMAKE_CXX_FLAGS "/O2 /FS /EHsc")
set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON) set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
endif(WIN32) endif(WIN32)
include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include/tkDNN) include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include/tkDNN)
README.md
+24 -15
View File
@@ -80,12 +80,13 @@ Results for COCO val 2017 (5k images), on RTX 2080Ti, with conf threshold=0.001
- [mAP demo](#map-demo) - [mAP demo](#map-demo)
- [Existing tests and supported networks](#existing-tests-and-supported-networks) - [Existing tests and supported networks](#existing-tests-and-supported-networks)
- [References](#references) - [References](#references)
- [tkDNN on Windows 10 (experimental)](#tkdnn-on-windows) - [tkDNN on Windows 10 (experimental)](#tkdnn-on-windows-10-experimental)
- [Dependencies-Windows](#dependencies-windows) - [Dependencies-Windows](#dependencies-windows)
- [Compiling tkDNN on Windows](#tkdnn-windows-compile) - [Compiling tkDNN on Windows](#compiling-tkdnn-on-windows)
- [Run the demo on Windows](#run-the-demo-on-windows) - [Run the demo on Windows](#run-the-demo-on-windows)
- [FP16 interference windows](#fp16-windows) - [FP16 inference windows](#fp16-inference-windows)
- [INT8 interference windows](#int8-windows) - [INT8 inference windows](#int8-inference-windows)
- [Known issues with tkDNN on Windows](#known-issues-with-tkdnn-on-windows)
@@ -362,26 +363,31 @@ This demo also creates a json file named ```net_name_COCO_res.json``` containing
| yolo4tiny | Yolov4 tiny <sup>9</sup> | [COCO 2017](http://cocodataset.org/) | 80 | 416x416 | [weights](https://cloud.hipert.unimore.it/s/iRnc4pSqmx78gJs/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 | 672x672 | [weights](https://cloud.hipert.unimore.it/s/BLPpiAigZJLorQD/download) | | yolo4x | Yolov4x-mish <sup>9</sup> | [COCO 2017](http://cocodataset.org/) | 80 | 672x672 | [weights](https://cloud.hipert.unimore.it/s/BLPpiAigZJLorQD/download) |
##tkDNN on Windows 10 (experimental) ### tkDNN on Windows 10 (experimental)
### Dependencies-Windows ### Dependencies-Windows
This branch should work on every NVIDIA GPU supported in windows with the following dependencies: This branch should work on every NVIDIA GPU supported in windows with the following dependencies:
* WINDOWS 10 1803 or HIGHER * WINDOWS 10 1803 or HIGHER
* CUDA 10.0 (Recommended CUDA 11.0 +) * CUDA 10.0 (Recommended CUDA 11.2 )
* CUDNN 7.6 (Recommended CUDNN 8.0.0 +) * CUDNN 7.6 (Recommended CUDNN 8.1.1 )
* TENSORRT 6.0.1 (Recommended TENSORRT 7.1 +) * TENSORRT 6.0.1 (Recommended TENSORRT 7.2.3.4 )
* OPENCV 3.4 (Recommended OPENCV 4.2.0 +) * OPENCV 3.4 (Recommended OPENCV 4.2.0 )
* MSVC 16.7 (Recommended MSVC 16.8/16.9) * MSVC 16.7
* YAML-CPP 0.5.2 * YAML-CPP
* EIGEN3 * EIGEN3
* 7ZIP (ADD TO PATH) * 7ZIP (ADD TO PATH)
* NINJA 1.10 * NINJA 1.10
All the above mentioned dependencies except 7ZIP can be installed using Microsoft's [VCPKG](https://github.com/microsoft/vcpkg.git) . 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 : After bootstrapping VCPKG the dependencies can be built and installed using the following command :
```vcpkg.exe install opencv4[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``` ```
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
```
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 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
@@ -411,7 +417,7 @@ Once the rt file has been successfully create,run the demo using the following c
``` ```
For general info on more demo paramters,check Run the demo section on top For general info on more demo paramters,check Run the demo section on top
### FP16 interference windows ### 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): This is an untested feature on windows.To run the object detection demo with FP16 interference follow the below steps(example with yolo4tiny):
``` ```
@@ -421,7 +427,7 @@ del /f yolo4tiny_fp16.rt
.\demo.exe yolo4tiny_fp16.rt ..\demo\yolo_test.mp4 .\demo.exe yolo4tiny_fp16.rt ..\demo\yolo_test.mp4
``` ```
### INT8 interference windows ### INT8 inference windows
To run object detection demo with INT8 (example with yolo4tiny): To run object detection demo with INT8 (example with yolo4tiny):
``` ```
set TKDNN_MODE=INT8 set TKDNN_MODE=INT8
@@ -433,10 +439,13 @@ del /f yolo4tiny_int8.rt # be sure to delete(or move) old tensorRT files
``` ```
### 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.
demo/demo/demo.cpp
+6 -1
View File
@@ -25,7 +25,12 @@ int main(int argc, char *argv[]) {
std::string net = "yolo4tiny_fp32.rt"; std::string net = "yolo4tiny_fp32.rt";
if(argc > 1) if(argc > 1)
net = argv[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) if(argc > 2)
input = argv[2]; input = argv[2];
char ntype = 'y'; char ntype = 'y';
include/tkDNN/DetectionNN.h
+1 -1
View File
@@ -17,7 +17,7 @@
#include "tkdnn.h" #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 #ifdef OPENCV_CUDACONTRIB
#include <opencv2/cudawarping.hpp> #include <opencv2/cudawarping.hpp>
include/tkDNN/pluginsRT/UpsampleRT.h
-2
View File
@@ -59,8 +59,6 @@ public:
tk::dnn::writeBUF(buf, c); tk::dnn::writeBUF(buf, c);
tk::dnn::writeBUF(buf, h); tk::dnn::writeBUF(buf, h);
tk::dnn::writeBUF(buf, w); tk::dnn::writeBUF(buf, w);
std::cout << "Upsample Serialization SIze" << getSerializationSize() << std::endl;
assert(buf == a + getSerializationSize()); assert(buf == a + getSerializationSize());
} }
include/tkDNN/pluginsRT/YoloRT.h
-1
View File
@@ -120,7 +120,6 @@ public:
tk::dnn::writeBUF(buf, tmp[j]); tk::dnn::writeBUF(buf, tmp[j]);
} }
} }
std::cout << getSerializationSize() << std::endl;
assert(buf == a + getSerializationSize()); assert(buf == a + getSerializationSize());
} }
scripts/download_validation.py
+2 -2
View File
@@ -24,7 +24,7 @@ file1 = open(".\\..\\demo\\all_labels.txt","a")
path1 = os.path.realpath(labelFolder) path1 = os.path.realpath(labelFolder)
for file in os.listdir(labelFolder): for file in os.listdir(labelFolder):
valTemp = path1 + "\\" + file valTemp = path1 + "\\" + file
valTemp = valTemp + " \n" valTemp = valTemp + '\n'
file1.write(valTemp) file1.write(valTemp)
file1.close() file1.close()
@@ -32,7 +32,7 @@ file2 = open(".\\..\\demo\\all_images.txt","a")
path2 = os.path.realpath(imageFolder) path2 = os.path.realpath(imageFolder)
for file in os.listdir(imageFolder): for file in os.listdir(imageFolder):
pathtemp = path2 + "\\" + file pathtemp = path2 + "\\" + file
pathtemp = pathtemp + " \n" pathtemp = pathtemp + '\n'
file2.write(pathtemp) file2.write(pathtemp)
file2.close() file2.close()
src/NetworkRT.cpp
+1 -1
View File
@@ -648,7 +648,7 @@ IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialDa
const char * buf = reinterpret_cast<const char*>(serialData),*bufCheck = buf; const char * buf = reinterpret_cast<const char*>(serialData),*bufCheck = buf;
std::string name(layerName); std::string name(layerName);
std::cout<<name<<std::endl; //std::cout<<name<<std::endl;
if(name.find("ActivationLeaky") == 0) { if(name.find("ActivationLeaky") == 0) {
ActivationLeakyRT *a = new ActivationLeakyRT(); ActivationLeakyRT *a = new ActivationLeakyRT();