This repo contains the official code of the project "Dynamic Residual Classifier for Class Incremental Learning" (ICCV2023). [Paper Link], [Supplementary].

First we recommend to create a conda environment with all the required packages by using the following command.

conda env create -f environment.yml

This command creates a conda environment named MAFDRC. You can activate the conda environment with the following command:

conda activate MAFDRC

In the following sections, we assume that you use this conda environment or you manually install the required packages.

Note that you may need to adapt the environment.yml/requirements.txt files to your infrastructure. The configuration of these files was tested on Linux Platform with a GPU (RTX3080 Ti).

If you see the following error, you may need to install a PyTorch package compatible with your infrastructure.

RuntimeError: No HIP GPUs are available or ImportError: libtinfo.so.5: cannot open shared object file: No such file or directory

For example if your infrastructure only supports CUDA == 11.1, you may need to install the PyTorch package using CUDA11.1.

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

We have implemented the pre-processing of CIFAR100, ImageNet100, ImageNet1000. When training on CIFAR100, this framework will automatically download it. When training on ImageNet100/ImageNet1000, you should specify the folder of your dataset in utils/data.py.

    def download_data(self):
        assert 0,"You should specify the folder of your dataset"
        train_dir = '[DATA-PATH]/train/'
        test_dir = '[DATA-PATH]/val/'

Testing:

• CIFAR100 B0 10 steps

  python main.py --config=mafdrc-cifar100.json --test True
  

Training:

• CIFAR100 B0 10 steps

  python main.py --config=mafdrc-cifar100.json
  

• ImageNet100 B0 10 steps

  python main.py --config=mafdrc-imagenet100.json
  

• ImageNet1000 B0 10 steps

  python main.py --config=mafdrc-imagenet1000.json 
  

Modifying "init_cls" and "increment" in mafdrc-[dataset].json for other CIL settings.

If you find this code useful, please kindly cite the following paper:

@article{drc-iccv,
  title={Dynamic Residual Classifier for Class Incremental Learning},
  author={Xiuwei Chen, Xiaobin Chang},
  booktitle = {IEEE International Conference on Computer Vision (ICCV)},
  year={2023}
}

This code is built on ECCV22-FOSTER.