Calculate the piecewise rotated/skew IOU of two lists of boxes.
Based on the custom op guide for TensorFlow
docker pull tensorflow/tensorflow:2.1.0-custom-op-gpu-ubuntu16 # so far there is no image for tensorflow-2.2.0
docker run -it -v ${PWD}:/working_dir -w /working_dir tensorflow/tensorflow:2.1.0-custom-op-gpu-ubuntu16
Within the docker container:
# Answer all questions with Y
./configure.sh
bazel build build_pip_pkg
bazel-bin/build_pip_pkg artifacts
You will find a python whl file in the artifacts folder.
To run the tests:
export PATH=/usr/local/cuda-10.1/bin/:/usr/local/cuda-10.1/lib64/:$PATH
make rotated_iou_test
There are some test to measure the performance under real world use cases:
export PATH=/usr/local/cuda-10.1/bin/:/usr/local/cuda-10.1/lib64/:$PATH
make benchmark
Some FAQ answers can be found with the original repo: