sua-choi / cms Goto Github PK

Download the datasets, ssb splits, and pretrained backbone by following the file structure below and set DATASET_ROOT={YOUR DIRECTORY} in config.py.

    DATASET_ROOT/
    ├── cifar100/
    │   ├── cifar-100-python\
    │   │   ├── meta/
    │       ├── ...
    ├── CUB_200_2011/
    │   ├── attributes/
    │   ├── ...
    ├── ...

    CMS/
    ├── data/
    │   ├── ssb_splits/
    ├── models/
    │   ├── dino_vitbase16_pretrain.pth
    ├── ...

Training

bash bash_scripts/contrastive_meanshift_training.sh

Example bash commands for training are as follows:

# GCD
python -m methods.contrastive_meanshift_training  \
            --dataset_name 'cub' \
            --lr 0.05 \
            --temperature 0.25 \
            --wandb 

# Inductive GCD
python -m methods.contrastive_meanshift_training  \
            --dataset_name 'cub' \
            --lr 0.05 \
            --temperature 0.25 \
            --inductive \
            --wandb

Evaluation

bash bash_scripts/meanshift_clustering.sh

Example bash command for evaluation is as follows. It will require changing model_name.

python -m methods.meanshift_clustering \
        --dataset_name 'cub' \
        --model_name 'cub_best' \

Results and checkpoints

Experimental results on GCD task.

	All	Old	Novel	Checkpoints
CIFAR100	82.3	85.7	75.5	link
ImageNet100	84.7	95.6	79.2	link
CUB	68.2	76.5	64.0	link
Stanford Cars	56.9	76.1	47.6	link
FGVC-Aircraft	56.0	63.4	52.3	link
Herbarium19	36.4	54.9	26.4	link

Experimental results on inductive GCD task.

	All	Old	Novel	Checkpoints
CIFAR100	80.7	84.4	65.9	link
ImageNet100	85.7	95.7	75.8	link
CUB	69.7	76.5	63.0	link
Stanford Cars	57.8	75.2	41.0	link
FGVC-Aircraft	53.3	62.7	43.8	link
Herbarium19	46.2	53.0	38.9	link

Citation

If you find our code or paper useful, please consider citing our paper:

  @inproceedings{choi2024contrastive,
    title={Contrastive Mean-Shift Learning for Generalized Category Discovery},
    author={Choi, Sua and Kang, Dahyun and Cho, Minsu},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
    year={2024}
  }

Related Repos

The codebase is largely built on Generalized Category Discovery and PromptCAL.

Acknowledgements

This work was supported by the NRF grant (NRF-2021R1A2C3012728 (50%)) and the IITP grants (2022-0-00113: Developing a Sustainable Collaborative Multi-modal Lifelong Learning Framework (45%), 2019-0-01906: AI Graduate School Program at POSTECH (5%)) funded by Ministry of Science and ICT, Korea.

cms's People

Contributors

Stargazers

Watchers

Forkers

dahyun-kang

cms's Issues

question of aircraft dataset

Excuse me, i download the provided weight of aircraft dataset, but the result is

Namespace(batch_size=128, num_workers=8, pretrain_path='./models', transform='imagenet', eval_funcs=['v2'], use_ssb_splits=True, model_name='aircraft_p', dataset_name='aircraft', epochs=20, feat_dim=768, num_clusters=None, inductive=False, k=8, alpha=0.5, image_size=224, train_classes=[0, 1, 2, 3, 4, 5, 10, 11, 14, 16, 17, 19, 21, 22, 23, 24, 27, 28, 29, 30, 33, 36, 37, 38, 39, 41, 43, 44, 45, 46, 47, 48, 52, 53, 56, 57, 58, 63, 64, 65, 66, 67, 71, 73, 76, 77, 79, 92, 95, 99], unlabeled_classes=[6, 7, 8, 9, 12, 13, 31, 32, 25, 26, 18, 20, 15, 78, 82, 51, 49, 50, 54, 55, 59, 60, 61, 68, 69, 70, 85, 86, 87, 88, 80, 81, 42, 84, 40, 90, 74, 75, 97, 98, 34, 35, 93, 94, 96, 72, 91, 83, 62, 89], num_labeled_classes=50, num_unlabeled_classes=50)
Using weights from /home/GEN/code/CMS/log/cms/log/aircraft_p/checkpoints/model_best.pt ...
/home/GEN/miniconda3/envs/torch1/lib/python3.10/site-packages/torchvision/transforms/transforms.py:329: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
  warnings.warn(
Predicted number of clusters:  92
100%|██████████████████████████████████████████████████████████████████████████████████████████████| 53/53 [00:34<00:00,  1.54it/s]
num clusters 100
Epoch 0, IMS unlabeled train ACC_v2: All 0.5353 | Old 0.6158 | New 0.4951
num clusters 92
Epoch 0, IMS unlabeled train ACC_v2: All 0.5373 | Old 0.6176 | New 0.4972
100%|██████████████████████████████████████████████████████████████████████████████████████████████| 53/53 [00:19<00:00,  2.74it/s]
num clusters 100
Epoch 1, IMS unlabeled train ACC_v2: All 0.5343 | Old 0.6170 | New 0.4930
num clusters 92
Epoch 1, IMS unlabeled train ACC_v2: All 0.5433 | Old 0.5984 | New 0.5157
100%|██████████████████████████████████████████████████████████████████████████████████████████████| 53/53 [00:19<00:00,  2.71it/s]
num clusters 100
Epoch 2, IMS unlabeled train ACC_v2: All 0.5393 | Old 0.6164 | New 0.5007
num clusters 92
Epoch 2, IMS unlabeled train ACC_v2: All 0.5461 | Old 0.5936 | New 0.5223
ACC with GT number of clusters: All 0.5393 | Old 0.6164 | New 0.5007
ACC with predicted number of clusters: All 0.5461 | Old 0.5936 | New 0.5223

the number is smaller than README (53 vs. 56).

And I also try to train the model by myself, but the number is also small than paper(CUB 66, SCARS 48, AIRCRAFT 51). I use the default hyperparameters on github, and no change to codes. I think maybe there are something wrong in my experiment, do you have any idea can help me fix the problem?

I look forward to your response at your earliest convenience.

About Clustering without the gt number of K

Thank you for the great work on GCD.
I would like to ask how to train the model on cub without the real k value. Which part of the code should I change?
Looking forward to your response and would like to thank you once again for your great work !

sua-choi / cms Goto Github PK

cms's Introduction

Contrastive Mean-Shift Learning for Generalized Category Discovery

Pohang University of Science and Technology (POSTECH)

[Paper] [Project page]

Environmnet installation

Datasets