MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition [Official, ICCV 2023, Paper] 🔥

Qihao Zhao^1,2, Chen Jiang¹, Wei Hu¹, Fan Zhang¹, Jun Liu²

1 Beijing University of Chemical Technology

2 Singapore University of Technology and Design

0.Citation

If you find our work inspiring or use our codebase in your research, please consider giving a star ⭐ and a citation.

@InProceedings{Zhao_2023_ICCV,
    author    = {Zhao, Qihao and Jiang, Chen and Hu, Wei and Zhang, Fan and Liu, Jun},
    title     = {MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {11597-11608}
}

1.training

(1) CIFAR100-LT

Training

• run:

  python train.py -c configs/config_cifar100_ir100_mdcs.json

Evaluate

• run:

  python test.py -r checkpoint_path

(2) ImageNet-LT

Training

• run such as resnext50 400 epochs:

  python train.py -c configs/config_imagenet_lt_resnext50_mdcs_e400.json

Evaluate

• run:

  python test.py -r checkpoint_path

(3) Places-LT

Training

• run:

  python train_places.py -c configs/config_places_lt_resnet152_mdcs.json

Evaluate

• run:

  python test_places.py -r checkpoint_path

(4) iNaturalist 2018

Training

• run :

  python train.py -c configs/config_iNaturalist_resnet50_mdcs.json

Evaluate

• run:

  python test.py -r checkpoint_path

2. Requirements

• To install requirements:

  pip install -r requirements.txt

• Run in linux (may have some problems in windows)

3. Datasets

(1) Four bechmark datasets

• Please download these datasets and put them to the /data file.
• ImageNet-LT and Places-LT can be found at here.
• iNaturalist data should be the 2018 version from here.
• CIFAR-100 will be downloaded automatically with the dataloader.

data
├── ImageNet_LT
│   ├── test
│   ├── train
│   └── val
├── CIFAR100
│   └── cifar-100-python
├── Place365
│   ├── data_256
│   ├── test_256
│   └── val_256
└── iNaturalist 
    ├── test2018
    └── train_val2018

(2) Txt files

• We provide txt files for test-agnostic long-tailed recognition for ImageNet-LT, Places-LT and iNaturalist 2018. CIFAR-100 will be generated automatically with the code.
• For iNaturalist 2018, please unzip the iNaturalist_train.zip.

  data_txt
  ├── ImageNet_LT
  │   ├── ImageNet_LT_test.txt
  │   ├── ImageNet_LT_train.txt
  │   └── ImageNet_LT_val.txt
  ├── Places_LT_v2
  │   ├── Places_LT_test.txt
  │   ├── Places_LT_train.txt
  │   └── Places_LT_val.txt
  └── iNaturalist18
  ├── iNaturalist18_train.txt
  ├── iNaturalist18_uniform.txt
  └── iNaturalist18_val.txt 

4. Pretrained models

• For the training on Places-LT, we follow previous methods and use the pre-trained ResNet-152 model.
• Please download the checkpoint. Unzip and move the checkpoint files to /model/pretrained_model_places/.

5. Acknowledgements

The mutli-expert framework is based on SADE and [RIDE](). Strong augmentations are based on NCL and PaCo.