The PyTorch Implementation for HFC

[ICCV-2023] Heterogeneous Forgetting Compensation for Class-Incremental Learning

Requirements

• torch>=1.7.0
• torchvision>=0.8.1
• timm==0.6.5
• continuum>=1.0.27
• numpy
• scikit-learn

Datasets

CIFAR100

You don't need to do anything before running the experiments on CIFAR100 dataset.

ImageNet100

Refer to ImageNet100_Split

ImageNet1000

Data preparation: download and extract ImageNet images from http://image-net.org/. The directory structure should be

│ILSVRC2012/
├──train/
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......
│  ├── ......
├──val/
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......
│  ├── ......

Experiments

You should pretrain the backbone ViT-B refer to https://github.com/facebookresearch/mae, then give the the path of model in option.py(--model_path).

Training

For exampler, if you wangt to run CILformer on CIFAR100 in the 10 steps setting:

Modify the path of dataset in './scripts/cifar/task10.sh'.

sh scripts/cifar/task10.sh

Results

The results of HFC will be written in './traning_log'.

Acknowledgement

Thanks for the great code base from https://github.com/DRSAD/iCaRL and https://github.com/arthurdouillard/dytox.

Citations

If you find this code is useful to your research, please consider citing the following paper.

@InProceedings{Dong_2023_ICCV,
    author    = {Dong, Jiahua and Liang, Wenqi and Cong, Yang and Sun, Gan},
    title     = {Heterogeneous Forgetting Compensation for Class-Incremental Learning},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {11742-11751}
}