/VSIAD

Learn to Recover Visible Color for Video Surveillance in a Day, ECCV 2020(oral),

Primary LanguagePythonMIT LicenseMIT

"Learn to Recover Visible Color for Video Surveillance in a Day" (ECCV2020 Oral)

This repository contains the code and dataset for the paper "Learn to Recover Visible Color for Video Surveillance in a Day" by Guangming Wu, Yinqiang Zheng, Zhiling Guo, et al.

Requirements

The source code have been valified on the following env:

  • NVIDIA GPU > 8GB
  • Ubuntu 16.04 LTS
  • Anaconda Python(3.6)
  • Pytorch > 1.0
  • OpenCV 3

VSIAD Dataset

Link => https://univtokyo-my.sharepoint.com/:u:/g/personal/9291958590_utac_u-tokyo_ac_jp/EZTYb_6LBWhEv1je1gzT8-wB1Dud678ewydOS_wM1YawdA?e=8Y37Kx
passwd => eccv2020

Pretrained model

checkppoint/
--res18ynet-VC24-RGB2LAB_iter_100000.pth                 => (SSN, SSM => 0)
--res18ynetsync-VC24-L2LAB_iter_100000.pth               => (SSN, L2LAB)
--res18ynetsync-VC24-RGB2LAB_iter_100000.pth             => (SSN, full model)
--res18ynetsync-VC24-RGB2LAB-nop_iter_100000.pth.pth     => (SSN, perceptual => 0)
--res18ynetsync-VC24-LAB2LAB_iter_100000.pth             => (SSN, LAB2LAB)
--res18ynetsyncGN-VC24-RGB2LAB_iter_100000.pth           => (SSN, IN2GN)

Usage

pending

Publication

G. Wu, Y. Zheng, Z. Guo, et al., "Learn to Recover Visible Color for Video Surveillance in a Day," European Conference on Computer Vision (ECCV), to appear, 2020. (oral presentation, acceptance rate: 2%)

Citation

@inproceedings{wu2020learn,
  title={Learn to Recover Visible Color for Video Surveillance in a Day},
  author={Wu, Guangming and Zheng, Yinqiang and Guo, Zhiling and Cai, Zekun and Shi, Xiaodan and Ding, Xin and Huang, Yifei and Guo, Yimin and Shibasaki, Ryosuke},
  booktitle={European Conference on Computer Vision},
  pages={495--511},
  year={2020},
  organization={Springer}
}

Contact

For any question, please contact

huster-wgm(.at.)csis.u-tokyo.ac.jp or
guangmingwu2010(.at.)gmail.com