I made Chainer implementation of Light Head RCNN. All codes are available on Github knorth55/chainer-light-head-rcnn. arXiv Github github.com

Use CUDA_VISIBLE_DEVICES environment variable as below. CUDA_VISIBLE_DEVICES=0,2,3 mpiexec -n 3 python train.py https://devtalk.nvidia.com/default/topic/509221/specifying-gpu-device-using-openmpi-mpirun/

I implemented Fully Convolutional Instance-aware Segmentation (FCIS) by Chainer. All codes including training scripts is on Github. I also made a explanation slide of FCIS on SlideShare. ArXiv Github SlideShare github.com