Hprc banner tamu.png


Revision as of 11:43, 20 March 2018 by Yangliu (talk | contribs) (Anaconda and Faster-RCNN_TF Packages)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search



This is an experimental Tensorflow implementation of Faster RCNN - a convnet for object detection with a region proposal network. For details about R-CNN please refer to the paper Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks by Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun.


Faster-RCNN_TF is open to all HPRC users.

There are restrictions as to which version (GPU/CPU) of Faster-RCNN_TF works on each cluster. Please note these restrictions in the following sections and plan your jobs accordingly.

Anaconda and Faster-RCNN_TF Packages

TAMU HPRC currently supports the user of Faster-RCNN_TF though the Anaconda modules. However, a user needs to install faster-rcnn_tf on their own following the instructions below:

[NetID@terra3 ~]$module load CUDA/8.0.44-GCC-system
[NetID@terra3 ~]$module load Anaconda/2-5.0.1
[NetID@terra3 ~]$source activate faster-r-cnn
(faster-r-cnn)[NetID@terrra3 ~] now you need edit the lib/make.sh used for installation of faster-rcnn_tf as the following:
TF_INC=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())')
TF_LIB=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_lib())')
cd roi_pooling_layer
if [ -d "$CUDA_PATH" ]; then
    nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc \
        -I$TF_INC -I$HEADER -L$TF_LIB -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CXXFLAGS --expt-relaxed-constexpr \
    g++ -std=c++11 -shared -ltensorflow_framework -o roi_pooling.so roi_pooling_op.cc \
         roi_pooling_op.cu.o -I $TF_INC -L$TF_LIB -I $HEADER -D GOOGLE_CUDA=1 -fPIC $CXXFLAGS \
         -lcudart -L $CUDA_PATH/lib
    g++ -std=c++11 -I$HEADER -L$TF_LIB -ltensorflow_framework -shared -o roi_pooling.so roi_pooling_op.cc \
         -I $TF_INC -fPIC $CXXFLAGS
cd ..

(faster-r-cnn) [netid@terra3 lib]$ now you can run 'make' to install.

After you install the Faster-RCNN_TF, you can run programs to access the package. But CUDA/8.0.44-GCC-system is not required to be loaded any more. That is, you only need do the followings to access the package:

[NetID@terra3 ~]$module load Anaconda/2-5.0.1
[NetID@terra3 ~]$source activate faster-r-cnn