Skip to content

wdjang/CTN

Repository files navigation


CTN


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.


Quick start

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.

  1. 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

  2. Pre-trained model can be downloaded from https://www.dropbox.com/s/gom5rv3mutrc0ju/CTN_iter_282816.caffemodel?dl=0

  3. Download sample data for demo https://www.dropbox.com/s/mj022oqbu415lvn/data.tar.gz?dl=0

  4. Please see 'demo_CTN.m'


LICENSE

This program is released with a research only license.

About

Convolutional Trident Network

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors