Run with TF 1.12.0

[UpCoder]

I successfully run the code with TF 1.12.0
You need run two sh instead of the make file.
In roi_pooling_layer, please run the sh file

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())')

echo $TF_INC

echo $TF_LIB

CUDA_PATH=/usr/local/cuda/

g++ -std=c++11 -c roi_pooling_op.cc -o roi_pooling_op.o -fPIC -I "/root/anaconda2/envs/ld_tf_base/lib/python3.6/site-packages/tensorflow/include" -O2

nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC -arch=sm_52

## if you install tf using already-built binary, or gcc version 4.x, uncomment the two lines below

#g++ -std=c++11 -shared -D_GLIBCXX_USE_CXX11_ABI=0 -o roi_pooling.so roi_pooling_op.cc \

# roi_pooling_op.cu.o -I $TF_INC -fPIC -lcudart -L $CUDA_PATH/lib64

# for gcc5-built tf

g++ -std=c++11 -shared -D_GLIBCXX_USE_CXX11_ABI=0 -o roi_pooling.so roi_pooling_op.cc roi_pooling_op.cu.o -I $TF_INC -I $TF_INC"/external/nsync/public" -L $TF_LIB -ltensorflow_framework -O2 -fPIC -lcudart -L $CUDA_PATH/lib64

cd ..

in matching_module run the sh file

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())')

echo $TF_INC

echo $TF_LIB

CUDA_PATH=/usr/local/cuda/

g++ -std=c++11 -c det_matching.cc -o det_matching.o -fPIC -I "/root/anaconda2/envs/ld_tf_base/lib/python3.6/site-packages/tensorflow/include" -O2

# nvcc -std=c++11 -c -o det_matching.o det_matching.cc -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC -arch=sm_52

## if you install tf using already-built binary, or gcc version 4.x, uncomment the two lines below

#g++ -std=c++11 -shared -D_GLIBCXX_USE_CXX11_ABI=0 -o roi_pooling.so roi_pooling_op.cc \

# roi_pooling_op.cu.o -I $TF_INC -fPIC -lcudart -L $CUDA_PATH/lib64

# for gcc5-built tf

g++ -std=c++11 -shared -D_GLIBCXX_USE_CXX11_ABI=0 -o det_matching.so det_matching.cc -I $TF_INC -I $TF_INC"/external/nsync/public" -L $TF_LIB -ltensorflow_framework -O2 -fPIC -lcudart -L $CUDA_PATH/lib64

cd ..

download the coco json annotation in the direction ProjectDir/data/coco/annotations

• instances_minival2014.json
• instances_train2014.json
• instances_val2014.json
• instances_valminusminival2014.json

Please train

python train.py --config=experiments/coco_person/conf.yaml

the result like this:

2020-08-17 03:18:11.800835  iter  59940   lr   0.0001   opt loss  1.90954     data loss normalized 0.0163604   unnormalized  1.76164
2020-08-17 03:18:11.940283  iter  59960   lr   0.0001   opt loss  6.11693     data loss normalized 0.032737   unnormalized  5.96905
2020-08-17 03:18:12.069166  iter  59980   lr   0.0001   opt loss 0.713293     data loss normalized 0.00841691   unnormalized 0.565408
2020-08-17 03:18:12.189150  iter  60000   lr   0.0001   opt loss  3.09831     data loss normalized 0.0365992   unnormalized  2.95046
2020-08-17 03:18:12.189251  starting validation
[1]
2020-08-17 03:18:27.687428  iter  60000   validation pass:   mAP  66.7   multiclass AP  66.7
/root/ld/PycharmProjects/gossipnet/outputs/gnet
wrote model to /root/ld/PycharmProjects/gossipnet/outputs/gnet-60000
Iteration   mAP   
     20000    64.0
     40000    65.9
     60000    66.7  (best)