Based on pytorch-faster-rcnn

CPU faster-RCNN

1. Deleted all training scripts
2. Wrote a new script for extracting all boxes from each image in a given folder
3. Compiled for CPUs (exclude all CUDA implementations). The original build files for CPU version are wrong, we corrected them.
4. Wrote a .sh script. Tried to be user-friendly.
5. The C-implemented layers are already compiled. Should work out-of-the-box.

Trained models can be downloaded from the RELEASE page of faster-RCNN-gpu

How to use

This is a CPU implementation for platforms without CUDA support. The run.py inside tools folder has the same usage as the GPU version