dylan-fan / chainer-faster-rcnn

Faster R-CNN with Chainer

Faster R-CNN

This is an experimental implementation of Faster R-CNN using Chainer based on Ross Girshick's py-faster-rcnn codes.

Requirement

• Python 2.7.6+, 3.4.3+, 3.5.1+

  • Chainer 1.9.1+
  • NumPy 1.9, 1.10, 1.11
  • Cython 0.23+
  • OpenCV 2.9+, 3.1+

Inference

1. Download pre-trained model

wget https://www.dropbox.com/s/2fadbs9q50igar8/VGG16_faster_rcnn_final.model?dl=0
mv VGG16_faster_rcnn_final.model?dl=0 VGG16_faster_rcnn_final.model

2. Build extensions

cd lib
python setup.py build_ext -i

3. Use forward.py

wget http://vision.cs.utexas.edu/voc/VOC2007_test/JPEGImages/004545.jpg

python forward.py --img_fn 004545.jpg --gpu 0

--gpu 0 turns on GPU. When you turn off GPU, use --gpu -1 or remove --gpu option.

To use forward.py with CPU, you have to apply this diff due to a known bug in Chainer: chainer/chainer#1273

Training

will be updated soon

Framework