Group-wise Contrastive Learning

This is the official code repository for the WACV 2024 paper "Group-wise Contrastive Bottleneck for Weakly-Supervised Visual Representation Learning".

Environment Setup

conda create --name clrc python=3.9
conda activate clrc

conda install cupy cudatoolkit=11.1 -c conda-forge
pip install -r requirements.txt

pip install .
pip install ./SupContrast

The SupContrast module is adapted from https://github.com/HobbitLong/SupContrast.

Datasets

Note: Images are not included in this repository. Please refer to the links below to download the images.

Name	Link
UT Zappos 50k	https://vision.cs.utexas.edu/projects/finegrained/utzap50k/
WIDER Attribtue	http://mmlab.ie.cuhk.edu.hk/projects/WIDERAttribute.html
CUB-200-2011	https://www.vision.caltech.edu/datasets/cub_200_2011/
ImageNet-100	https://image-net.org/download.php
Fitzpatrick17k	https://github.com/mattgroh/fitzpatrick17k

Pretraining

Configuration files for different datasets and pretrain methods can be found in the config/pretrain folder. An example of pretraining on UT Zappos 50k:

python main_pretrain.py \
    utzap-attribute \
    clrc \
    --config_file config/pretrain/utzap/clrc.json

Evaluation

Configuration files for different datasets can be found in the config/downstream folder. An example of linear evaluation on UT Zappos 50k:

python main_finetune.py \
    utzap \
    classifier \
    --backbone_weights model/pretrain/utzap/clrc/lightning_logs/version_0/checkpoints/epoch=999-step=272999.ckpt \
    --config_file config/downstream/utzap.json

Citation

@InProceedings{Yap_2024_WACV,
    author    = {Yap, Boon Peng and Ng, Beng Koon},
    title     = {Group-Wise Contrastive Bottleneck for Weakly-Supervised Visual Representation Learning},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2024},
    pages     = {2246-2255}
}