Merge pull request jwyang#189 from kingstarcraft/windows

Porting this code to Windows

Author: jwyang

.gitignore

@@ -1,3 +1,17 @@
 data/*
 *.pyc
 *~
+/images - Copy
+*.pyd
+*.ipch
+*.tlog
+*.deps
+*.cache
+*.obj
+*.log
+*.pdb
+/vgg16_caffe.pth
+/windows/.vs/faster_rcnn/v15
+*.dll
+*.lib
+/.vs/FasterRCNN/v15

FasterRCNN.pyproj

@@ -0,0 +1,23 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio 15
+VisualStudioVersion = 15.0.27703.2026
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{888888A0-9F3D-457C-B088-3A5042F75D52}") = "FasterRCNN", "FasterRCNN.pyproj", "{331BAC87-6D9D-4023-954A-0A35943A3358}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|Any CPU = Debug|Any CPU
+		Release|Any CPU = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{331BAC87-6D9D-4023-954A-0A35943A3358}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{331BAC87-6D9D-4023-954A-0A35943A3358}.Release|Any CPU.ActiveCfg = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+	GlobalSection(ExtensibilityGlobals) = postSolution
+		SolutionGuid = {4969271A-7E02-44A8-B9CD-00F0B4151CC0}
+	EndGlobalSection
+EndGlobal

FasterRCNN.sln

@@ -149,8 +149,10 @@ Install all the python dependencies using pip:
 pip install -r requirements.txt
 ```
 
-Compile the cuda dependencies using following simple commands:
-
+if bulid on windows, please click [README.md](./windows/README.md).
+   
+else compile the cuda dependencies using following simple commands:
+   
 ```
 cd lib
 sh make.sh

README.md

@@ -55,13 +55,13 @@ def parse_args():
                       default='cfgs/vgg16.yml', type=str)
   parser.add_argument('--net', dest='net',
                       help='vgg16, res50, res101, res152',
-                      default='res101', type=str)
+                      default='vgg16', type=str)
   parser.add_argument('--set', dest='set_cfgs',
                       help='set config keys', default=None,
                       nargs=argparse.REMAINDER)
   parser.add_argument('--load_dir', dest='load_dir',
                       help='directory to load models',
-                      default="/srv/share/jyang375/models")
+                      default="F:/Pascal/faster_rcnn_pytorch")
   parser.add_argument('--image_dir', dest='image_dir',
                       help='directory to load images for demo',
                       default="images")

demo.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-CUDA_PATH=/usr/local/cuda/
+CUDA_PATH="A:/NVIDIA/GPU Computing Toolkit/CUDA/v9.1"
 
 python setup.py build_ext --inplace
 rm -rf build
@@ -15,8 +15,8 @@ CUDA_ARCH="-gencode arch=compute_30,code=sm_30 \
 # compile NMS
 cd model/nms/src
 echo "Compiling nms kernels by nvcc..."
-nvcc -c -o nms_cuda_kernel.cu.o nms_cuda_kernel.cu \
-	 -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
+#nvcc -c -o nms_cuda_kernel.lib nms_cuda_kernel.cu \
+#     -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
 
 cd ../
 python build.py
@@ -25,25 +25,25 @@ python build.py
 cd ../../
 cd model/roi_pooling/src
 echo "Compiling roi pooling kernels by nvcc..."
-nvcc -c -o roi_pooling.cu.o roi_pooling_kernel.cu \
-	 -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
+#nvcc -c -o roi_pooling.lib roi_pooling_kernel.cu \
+#     -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
 cd ../
 python build.py
 
 # compile roi_align
 cd ../../
 cd model/roi_align/src
 echo "Compiling roi align kernels by nvcc..."
-nvcc -c -o roi_align_kernel.cu.o roi_align_kernel.cu \
-	 -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
+#nvcc -c -o roi_align_kernel.lib roi_align_kernel.cu \
+#	  -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
 cd ../
 python build.py
 
 # compile roi_crop
 cd ../../
 cd model/roi_crop/src
 echo "Compiling roi crop kernels by nvcc..."
-nvcc -c -o roi_crop_cuda_kernel.cu.o roi_crop_cuda_kernel.cu \
-	 -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
+#nvcc -c -o roi_crop_cuda_kernel.lib roi_crop_cuda_kernel.cu \
+#     -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CUDA_ARCH
 cd ../
 python build.py

images/img1_det.jpg

@@ -6,7 +6,10 @@
 def _import_symbols(locals):
     for symbol in dir(_lib):
         fn = getattr(_lib, symbol)
-        locals[symbol] = _wrap_function(fn, _ffi)
+        if callable(fn):
+            locals[symbol] = _wrap_function(fn, _ffi)
+        else:
+            locals[symbol] = fn
         __all__.append(symbol)
 
 _import_symbols(locals())

images/img1_det_res101.jpg

@@ -1,25 +1,36 @@
 from __future__ import print_function
 import os
+import sys
 import torch
 from torch.utils.ffi import create_extension
 
 #this_file = os.path.dirname(__file__)
+torch_root = os.path.join(os.path.dirname(sys.executable),
+                          'Lib/site-packages/torch/lib')
+cuda_root = os.environ['CUDA_PATH']
 
 sources = []
 headers = []
 defines = []
+include_dirs = []
 with_cuda = False
 
+this_file = os.path.dirname(os.path.realpath(__file__))
+print(this_file)
+
 if torch.cuda.is_available():
     print('Including CUDA code.')
-    sources += ['src/nms_cuda.c']
+    sources += ['src/nms_cuda.cpp']
     headers += ['src/nms_cuda.h']
+    include_dirs += [os.path.join(cuda_root,"include"), os.path.join(torch_root,'include')]
     defines += [('WITH_CUDA', None)]
     with_cuda = True
 
-this_file = os.path.dirname(os.path.realpath(__file__))
-print(this_file)
-extra_objects = ['src/nms_cuda_kernel.cu.o']
+
+extra_objects =  ['src/nms_cuda_kernel.lib']
+extra_objects += [os.path.join(torch_root,'ATen.lib'),
+                  os.path.join(cuda_root,'lib/x64/cudart.lib'),
+                  os.path.join(torch_root,'_C.lib')]
 extra_objects = [os.path.join(this_file, fname) for fname in extra_objects]
 print(extra_objects)
 
@@ -30,7 +41,8 @@
     define_macros=defines,
     relative_to=__file__,
     with_cuda=with_cuda,
-    extra_objects=extra_objects
+    extra_objects=extra_objects,
+    include_dirs=include_dirs
 )
 
 if __name__ == '__main__':

images/img2.jpg

@@ -1,5 +1,4 @@
 // int nms_cuda(THCudaTensor *keep_out, THCudaTensor *num_out,
 //             THCudaTensor *boxes_host, THCudaTensor *nms_overlap_thresh);
-
 int nms_cuda(THCudaIntTensor *keep_out, THCudaTensor *boxes_host,
              THCudaIntTensor *num_out, float nms_overlap_thresh);

images/img2_det.jpg

@@ -6,7 +6,10 @@
 def _import_symbols(locals):
     for symbol in dir(_lib):
         fn = getattr(_lib, symbol)
-        locals[symbol] = _wrap_function(fn, _ffi)
+        if callable(fn):
+            locals[symbol] = _wrap_function(fn, _ffi)
+        else:
+            locals[symbol] = fn
         __all__.append(symbol)
 
 _import_symbols(locals())

images/img2_det_res101.jpg

@@ -1,13 +1,17 @@
 from __future__ import print_function
 import os
+import sys
 import torch
 from torch.utils.ffi import create_extension
 
-sources = ['src/roi_align.c']
+
+torch_root = os.path.join(os.path.dirname(sys.executable),
+                          'Lib/site-packages/torch/lib')
+cuda_root = os.environ['CUDA_PATH']
+
+sources = ['src/roi_align.cpp']
 headers = ['src/roi_align.h']
-extra_objects = []
-#sources = []
-#headers = []
+include_dirs = []
 defines = []
 with_cuda = False
 
@@ -16,13 +20,18 @@
 
 if torch.cuda.is_available():
     print('Including CUDA code.')
-    sources += ['src/roi_align_cuda.c']
+    sources += ['src/roi_align_cuda.cpp']
     headers += ['src/roi_align_cuda.h']
     defines += [('WITH_CUDA', None)]
     with_cuda = True
-    
-    extra_objects = ['src/roi_align_kernel.cu.o']
-    extra_objects = [os.path.join(this_file, fname) for fname in extra_objects]
+    include_dirs += [os.path.join(cuda_root,"include"), 
+                     os.path.join(torch_root,'include')]
+extra_objects = ['src/roi_align_kernel.lib']
+extra_objects = [os.path.join(this_file, fname) for fname in extra_objects]
+extra_objects += [os.path.join(torch_root,'ATen.lib'),
+                  os.path.join(cuda_root,'lib/x64/cudart.lib'),
+                  os.path.join(torch_root,'_C.lib')]
+print(extra_objects)
 
 ffi = create_extension(
     '_ext.roi_align',
@@ -31,7 +40,8 @@
     define_macros=defines,
     relative_to=__file__,
     with_cuda=with_cuda,
-    extra_objects=extra_objects
+    extra_objects=extra_objects,
+    include_dirs=include_dirs
 )
 
 if __name__ == '__main__':

images/img3.jpg

@@ -13,7 +13,7 @@ void ROIAlignBackwardCpu(const float* top_diff, const float spatial_scale, const
                      const int aligned_height, const int aligned_width, const float * bottom_rois,
                      float* top_data);
 
-int roi_align_forward(int aligned_height, int aligned_width, float spatial_scale,
+extern"C" __declspec(dllexport) int roi_align_forward(int aligned_height, int aligned_width, float spatial_scale,
                      THFloatTensor * features, THFloatTensor * rois, THFloatTensor * output)
 {
     //Grab the input tensor
@@ -44,7 +44,7 @@ int roi_align_forward(int aligned_height, int aligned_width, float spatial_scale
     return 1;
 }
 
-int roi_align_backward(int aligned_height, int aligned_width, float spatial_scale,
+extern"C" __declspec(dllexport) int roi_align_backward(int aligned_height, int aligned_width, float spatial_scale,
                        THFloatTensor * top_grad, THFloatTensor * rois, THFloatTensor * bottom_grad)
 {
     //Grab the input tensor

images/img3_det.jpg

@@ -1,10 +1,11 @@
 #include <THC/THC.h>
 #include <math.h>
+#include <ATen/ATen.h>
 #include "roi_align_kernel.h"
 
-extern THCState *state;
+THCState *state = at::globalContext().thc_state;
 
-int roi_align_forward_cuda(int aligned_height, int aligned_width, float spatial_scale,
+extern"C" __declspec(dllexport) int roi_align_forward_cuda(int aligned_height, int aligned_width, float spatial_scale,
                         THCudaTensor * features, THCudaTensor * rois, THCudaTensor * output)
 {
     // Grab the input tensor
@@ -39,7 +40,7 @@ int roi_align_forward_cuda(int aligned_height, int aligned_width, float spatial_
     return 1;
 }
 
-int roi_align_backward_cuda(int aligned_height, int aligned_width, float spatial_scale,
+extern"C" __declspec(dllexport) int roi_align_backward_cuda(int aligned_height, int aligned_width, float spatial_scale,
                         THCudaTensor * top_grad, THCudaTensor * rois, THCudaTensor * bottom_grad)
 {
     // Grab the input tensor

images/img3_det_res101.jpg

@@ -6,7 +6,10 @@
 def _import_symbols(locals):
     for symbol in dir(_lib):
         fn = getattr(_lib, symbol)
-        locals[symbol] = _wrap_function(fn, _ffi)
+        if callable(fn):
+            locals[symbol] = _wrap_function(fn, _ffi)
+        else:
+            locals[symbol] = fn
         __all__.append(symbol)
 
 _import_symbols(locals())

images/img4_det.jpg

@@ -1,26 +1,38 @@
 from __future__ import print_function
 import os
+import sys
 import torch
 from torch.utils.ffi import create_extension
 
 #this_file = os.path.dirname(__file__)
+torch_root = os.path.join(os.path.dirname(sys.executable),
+                          'Lib/site-packages/torch/lib')
+cuda_root = os.environ['CUDA_PATH']
 
-sources = ['src/roi_crop.c']
+sources = ['src/roi_crop.cpp']
 headers = ['src/roi_crop.h']
+include_dirs = []
 defines = []
 with_cuda = False
 
+this_file = os.path.dirname(os.path.realpath(__file__))
+print(this_file)
+
 if torch.cuda.is_available():
     print('Including CUDA code.')
-    sources += ['src/roi_crop_cuda.c']
+    sources += ['src/roi_crop_cuda.cpp']
     headers += ['src/roi_crop_cuda.h']
+    include_dirs += [os.path.join(cuda_root,"include"), 
+                     os.path.join(torch_root,'include')]
     defines += [('WITH_CUDA', None)]
     with_cuda = True
 
-this_file = os.path.dirname(os.path.realpath(__file__))
-print(this_file)
-extra_objects = ['src/roi_crop_cuda_kernel.cu.o']
+extra_objects = ['src/roi_crop_cuda_kernel.lib']
 extra_objects = [os.path.join(this_file, fname) for fname in extra_objects]
+extra_objects += [os.path.join(torch_root,'ATen.lib'),
+                  os.path.join(cuda_root,'lib/x64/cudart.lib'),
+                  os.path.join(torch_root,'_C.lib')]
+print(extra_objects)
 
 ffi = create_extension(
     '_ext.roi_crop',
@@ -29,7 +41,8 @@
     define_macros=defines,
     relative_to=__file__,
     with_cuda=with_cuda,
-    extra_objects=extra_objects
+    extra_objects=extra_objects,
+    include_dirs=include_dirs
 )
 
 if __name__ == '__main__':