This project provides the code and results for 'Effective Online Refinement for Video Object Segmentation', MTAP 2019. Paper link Homepege

Our code is implemented based on the OSVOS in Caffe. You can first install the 'caffe-osvos' according to the OSVOS.

We provide the one result of our 4 runs results here: DAVIS Results and Youtube Results.

1. Download the 20 trained model from here (1.2 G) and put it under models/DAVIS_github/.
2. Edit in file set_params_DAVIS_github.m the parameters of the code (eg. useGPU, gpu_id, etc.).
3. Run test_DAVIS_github.m.

    @ARTICLE{Li_2019_MTAP,
            author = {Li, Gongyang and Liu, Zhi and Zhou, Xiaofei},
            title = {Effective Online Refinement for Video Object Segmentatio},
            journal = {Multimedia Tools and Applications},
            year = {2019},
            volume = {78},
            number = {23},
            pages = {33617-33631},
            month = {Dec.},}

If you encounter any problems with the code, want to report bugs, etc.

Please contact me at lllmiemie@163.com or ligongyang@shu.edu.cn.