Code written by Won-Dong Jang Contact: Won-Dong Jang, wdjang@mcl.korea.ac.kr
If you want to use this software, please cite:
Won-Dong Jang and Chang-Su Kim, "Online Video Object Segmentation via Convolutional Trident Network," in Proc. IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, pp. 5849-5858, Jul. 2017.
This software was developed under 64-bit Linux with Matlab R2016b. There is no guarantee it will run on other operating systems or Matlab versions.
-
This code does not include a code for the optical flow estimation. Please download the codes at the below web-sites (Accurate Ver.) https://github.com/YinlinHu/CPM (Fast Ver.) https://github.com/tikroeger/OF_DIS
-
Pre-trained model can be downloaded from https://www.dropbox.com/s/gom5rv3mutrc0ju/CTN_iter_282816.caffemodel?dl=0
-
Download sample data for demo https://www.dropbox.com/s/mj022oqbu415lvn/data.tar.gz?dl=0
-
Please see 'demo_CTN.m'
- Pre-computed results on DAVIS dataset can be downloaded from https://www.dropbox.com/s/8dw5erk3zajcnh0/2017_CVPR_WDJANG_CTN_DAVIS.zip?dl=0
This program is released with a research only license.