faster rcnn tensorflow版本遇到的问题记录

1. fatal error: nsync_cv.h: No such file or directory

• a. 找到~/tensorflow/venvpy27/lib/python2.7/site-packages/tensorflow/include/tensorflow/core/platform/default文件夹下的mutex.h文件

将

#include "nsync_cv.h"
#include "nsync_mu.h"

改为

#include "external/nsync/public/nsync_cv.h"
#include "external/nsync/public/nsync_mu.h"

• b. 修改faster rcnn文件夹中lib文件夹下的make.sh文件为（注意空格）：

TF_INC=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())')
TF_LIB=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_lib())')
NSYNC_INC=$TF_INC"/external/nsync/public"
$TF_INC"/home/lbz/.conda/envs/py2/lib/python2.7/site-packages/tensorflow/contrib/makefile/downloads/nsync/public"

CUDA_PATH=/usr/local/cuda/
CXXFLAGS=''

if [[ "$OSTYPE" =~ ^darwin ]]; then
    CXXFLAGS+='-undefined dynamic_lookup'
fi

cd roi_pooling_layer

if [ -d "$CUDA_PATH" ]; then
    nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc \
        -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CXXFLAGS \
        -arch=sm_37

#    g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc \
#        roi_pooling_op.cu.o -I $TF_INC  -D GOOGLE_CUDA=1 -fPIC $CXXFLAGS \
#        -lcudart -L $CUDA_PATH/lib64
        g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc -D_GLIBCXX_USE_CXX11_ABI=0 \
                roi_pooling_op.cu.o -I $TF_INC -L $TF_LIB -ltensorflow_framework -D GOOGLE_CUDA=1 \
                -fPIC $CXXFLAGS -lcudart -L $CUDA_PATH/lib64
else
    g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc \
        -I $TF_INC -fPIC $CXXFLAGS
fi

cd ..

#cd feature_extrapolating_layer

#nvcc -std=c++11 -c -o feature_extrapolating_op.cu.o feature_extrapolating_op_gpu.cu.cc \
#    -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC -arch=sm_50

#g++ -std=c++11 -shared -o feature_extrapolating.so feature_extrapolating_op.cc \
#    feature_extrapolating_op.cu.o -I $TF_INC -fPIC -lcudart -L $CUDA_PATH/lib64
#cd ..

2. undefined symbol: _ZTIN10tensorflow8OpKernelE

解决：在每个 g++ 中加入  -L $TF_LIB

参考链接

1. fatal error: nsync_cv.h: No such file or directory

2.运行tensorflow版的fasterRCNN遇到的问题总结