1.配置環(huán)境
windows10 + VisualStudio2017community + YOLOV3
cuda10.0.130_411.31
cunn-10.0-v7.6.5.32
opencv3.2
常規(guī)
windows sdk版本 10.0.19041.0（10.0.17763.0）
平臺工具集 Visual Studio 2017(V141)
C++ ---常規(guī)
(CudaToolkitIncludeDir);
(cudnn)\include
VC++ 目錄
----可執(zhí)行文件目錄
D:\opencv\opencv\build\include;D:\opencv\opencv\build\include\opencv;D:\opencv\opencv\build\include\opencv2;(IncludePath)
----庫目錄
D:\opencv\opencv\build\x64\vc14\lib;$(LibraryPath)
鏈接器---常規(guī)---附加庫目錄
D:\opencv\opencv\build\x64\vc14\lib;(PlatformName);
(cudnn)\lib\x64;....\3rdparty\pthreads\lib;opencv_world320.lib;%(AdditionalLibraryDirectories)
鏈接器---輸入---附加依賴項
pthreadVC2.lib
cublas.lib
opencv_world320.lib
opencv_world320d.lib
curand.lib
cudart.lib
2.使用VS2017進行編譯
3.訓練
1.修改coco.data, coco.names
2.需要修改yolov3.cfg 一個是將test注釋掉袋倔，train注釋取消详恼，將classes類別數設置正確，卷積核filter個數設置為3*（類別數+1+4）
darknet.exe detector train data/coco.data cfg/yolov3.cfg darknet53.conv.74
4.預測
darknet.exe detector test data/coco.data cfg/yolov3.cfg backup/yolov3_final.weights sample.jpg
5.測試
darknet.exe detector vaild data/coco.data cfg/yolov3.cfg backup/yolov3_final.weights
6.視頻demo
darknet.exe detector demo data/coco.data cfg/yolov3.cfg backup/yolov3_final.weights "sample.mp4"
參考鏈接：
Windows10+YOLOV3+VisualStudio2017最新版本超詳細過程
CUDA10.1配置VS2017
cudnn下載鏈接
cuda下載鏈接
VS2017永久配置openCV
