This project has moved to ChainerCV.

chainer-light-head-rcnn - Light Head RCNN

This is Chainer implementation of Light-Head R-CNN: In Defense of Two-Stage Object Detector.

Original TensorFlow repository is zengarden/light_head_rcnn.

Requirement

• CuPy
• Chainer
• ChainerCV
• OpenCV

Additional Requirement

• For COCO Dataset class
  • Cython
  • pycocotools
• For training
  • OpenMPI with CUDA
  • NCCL
  • ChainerMN

Evaluation Score

Implementation	mAP@0.5:0.95/all	mAP@0.5/all	mAP@0.75/all	mAP:0.5:0.95/small	mAP:0.5:0.95/medium	mAP:0.5:0.95/large
Original	0.400	0.621	0.429	0.225	0.446	0.540
Ours	0.391	0.607	0.419	0.212	0.428	0.541

Installation

We recommend to use Anacoda.

# Requirement installation
conda create -n light-head-rcnn python=3.6
source activate light-head-rcnn
pip install opencv-python
pip install cupy

# Installation
git clone https://github.com/knorth55/chainer-light-head-rcnn.git
cd chainer-light-head-rcnn/
pip install -e .

Inference

cd examples/
python demo.py <imagepath> --gpu <gpu>

Training

cd examples/
mpiexec -n <n_gpu> python train_multi.py

LICENSE

MIT LICENSE