Weakly Supervised Contrastive Learning (ICCV2021)

This repository contains PyTorch evaluation code, training code and pretrained models for WCL.

For details see Weakly Supervised Contrastive Learning by Mingkai Zheng, Fei Wang, Shan You, Chen Qian, Changshui Zhang, Xiaogang Wang and Chang Xu

Reproducing

To run the code, you probably need to change the Dataset setting (dataset/imagenet.py), and Pytorch DDP setting (util/dist_init.py) for your own server enviroments.

The distribued training of this code is base on slurm enviroments, we have provide the training scrips under the script folder.

In this code, we adopt a hidden dimension of 4096 and output dimension 256 for the projection head (we use 2048 and 128 in our paper) since we found the performance can be further improved a little bit.

	Arch	BatchSize	Epochs	Linear Eval	Linear Eval (Paper)	Download
WCL	ResNet50	4096	100	68.5 %	68.1 %	wcl-100.pth
WCL	ResNet50	4096	200	70.5 %	70.3 %	wcl-200.pth

If you want to test the pretained model, please download the weights from the link above, and move it to the checkpoints folder (create one if you don't have .checkpoints/ directory). The evaluation scripts also has been provided in script/train.sh

Citation

If you find that wcl interesting and help your research, please consider citing it:

@InProceedings{Zheng_2021_ICCV,
    author    = {Zheng, Mingkai and Wang, Fei and You, Shan and Qian, Chen and Zhang, Changshui and Wang, Xiaogang and Xu, Chang},
    title     = {Weakly Supervised Contrastive Learning},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {10042-10051}
}