Skip to content

Latest commit

 

History

History
53 lines (45 loc) · 1.37 KB

README.md

File metadata and controls

53 lines (45 loc) · 1.37 KB
Raw

CoKe

PyTorch Implementation for Our CVPR'22 Paper: "Unsupervised Visual Representation Learning by Online Constrained K-Means"

Requirements

  • Python 3.8
  • PyTorch 1.6

Usage:

CoKe with single view

sh run_single_view.sh

CoKe with two views

sh run_double_view.sh

CoKe with multi-view

sh run_multi_view.sh

Linear evaluation on ImageNet

sh run_lincls.sh 0.2 coke_single_view.pth.tar

sh run_lincls.sh 0.4 coke_multi_view.pth.tar

Models:

model epochs crop acc on ImageNet link
CoKe 1000 1x224 72.5 Google Drive
CoKe 1000 2x224 74.9 Google Drive
CoKe 800 2x224+6x96 76.4 Google Drive

Citation

If you use the package in your research, please cite our paper:

@inproceedings{qian2022coke,
  author    = {Qi Qian and
               Yuanhong Xu and
               Juhua Hu and
               Hao Li and
               Rong Jin},
  title     = {Unsupervised Visual Representation Learning by Online Constrained K-Means},
  booktitle = {{IEEE} Conference on Computer Vision and Pattern Recognition, {CVPR} 2022},
  year      = {2022}
}