FCIS

Fully Convolutional Instance-aware Semantic Segmentation

(Previous name: translation-aware fully convolutional instance segmentation)

Yi Li*, Haozhi Qi*, Jifeng Dai, Xiangyang Ji, Yichen Wei

(* Equal contribution. Work was done when Yi Li and Haozhi Qi were interns at MSRA)

Introduction

This is the repository for Fully Convolutional Instance-aware Semantic Segmentation (FCIS), which is the winning entry of COCO segmentation challenge 2016. A arxiv tech report describing FCIS is available here

If you find FCIS useful for your research, please consider citing

@article{liang2015proposal, title={Fully Convolutional Instance-aware Semantic Segmentatio}, author={Li, Yi and Qi, Haozhi and Dai, Jifeng and Ji, Xiangyang and Wei, Yichen}, journal={arXiv preprint arXiv:1611.07709}, year={2016} }

<img src='data/readme_img/COCO_test2015_000000045082.png', width='300'> <img src='data/readme_img/COCO_test2015_000000186924.png', width='300'>

<img src='data/readme_img/COCO_test2015_000000174958.png', width='300'> <img src='data/readme_img/COCO_test2015_000000172816.png', width='300'>

Resources

0. Visual results on the first 5k images from COCO test set: OneDrive
1. Slides in ImageNet ILSVRC and COCO workshop 2016: OneDrive
2. Code of FCIS is coming soon!