DCCL

Offical implementation of our Dynamic Conceptional Contrastive Learning for Generalized Category Discovery in CVPR2023 (arXiv) by Nan Pu, Zhun Zhong, Nicu Sebe.

Abstract

Generalized category discovery (GCD) is a recently proposed open-world problem, which aims to automatically cluster partially labeled data. The main challenge is that the unlabeled data contain instances that are not only from known categories of the labeled data but also from novel categories. This leads traditional novel category discovery (NCD) methods to be incapacitated for GCD, due to their assumption of unlabeled data are only from novel categories. One effective way for GCD is applying selfsupervised learning to learn discriminate representation for unlabeled data. However, this manner largely ignores underlying relationships between instances of the same concepts (e.g., class, super-class, and sub-class), which results in inferior representation learning. In this paper, we propose a Dynamic Conceptional Contrastive Learning (DCCL) framework, which can effectively improve clustering accuracy by alternately estimating underlying visual conceptions and learning conceptional representation. In addition, we design a dynamic conception generation and update mechanism, which is able to ensure consistent conception learning and thus further facilitate the optimization of DCCL. Extensive experiments show that DCCL achieves new state-of-the-art performances on six generic and fine-grained visual recognition datasets, especially on fine-grained ones. For example, our method significantly surpasses the best competitor by 16.2% on the new classes for the CUB-200 dataset.

Requirements

• Python 3.8
• Pytorch 1.10.0
• torchvision 0.11.1

pip install -r requirements.txt

Datasets

In our experiments, we use generic image classification datasets including CIFAR-10/100 and ImageNet.

We also use fine-grained image classification datasets including CUB-200, Stanford-Cars, and Oxford-Pet.

Pretrained Checkpoints

Our model is initialized with the parameters pretrained by DINO on ImageNet. The DINO checkpoint of ViT-B-16 is available at here.

Training and Evaluation Instructions

Step 1. Set config

Set the path of datasets and the directory for saving outputs in config.py.

Step 2. Train and Test on CUB200 dataset

python G0_CUB200.py

Experiments on Other datasets

For experiments on other datasets, please modify parser.add_argument('--dataset_name', type=str, default='cub', help='options: imagenet_100,cifar10, cifar100, scars') in the demo code and refer to the paper for other hyperparameters' setting.

Results

Results of our method are reported as below. | Datasets | All | Old | New | |:------------|:--------:|:---------:|:---------:| | CIFAR10 | 96.3| 96.5| 96.9 | | CIFAR100 | 75.3 |76.8 |70.2 | | ImageNet-100 | 80.5| 90.5| 76.2 | | CUB-200 | 63.5 | 60.8 | 64.9 | | Stanford-Cars | 43.1| 55.7 | 36.2 | | Oxford-Pet | 88.1| 88.2 | 88.0 |

Citation

If you find this repo useful for your research, please consider citing our paper:

@inproceedings{pu2023dynamic,
  title={Dynamic Conceptional Contrastive Learning for Generalized Category Discovery},
  author={Pu, Nan and Zhong, Zhun and Sebe, Nicu},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={7579--7588},
  year={2023}
}

Acknowledgement

This project is modified from https://github.com/YiXXin/XCon. Thanks for their nice work.