Confidence Self-Calibration for Multi-Label Class-Incremental Learning

PyTorch code for the ECCV 2024 paper:
Confidence Self-Calibration for Multi-Label Class-Incremental Learning
Kaile Du, Yifan Zhou, Fan Lyu, Yuyang Li, Chen Lu, and Guangcan Liu
The 18th European Conference on Computer Vision ECCV 2024

Setup

To set up the environment and install the necessary dependencies, follow the steps below:

1. Install Anaconda from here.
2. Create a conda environment with Python 3.7. Example: conda create --name CSC python=3.7
3. Activate the conda environment: conda activate CSC
4. Install the required packages from requirements.txt: pip install -r requirements.txt

Datasets and Pretrained Model

Ensure you have the following datasets and pretrained models available:

• MS-COCO 2014: Download the dataset and place it in ./src/datasets/MSCOCO.
• PASCAL VOC 2007: Download the dataset and place it in ./src/datasets/VOC2007.
• TResNetM pretrained on ImageNet 21k: Download the pretrained model from here and rename it as tresnet_m_224_21k.pth. Place it in ./src/pretrained_tresnet.

Data Partitioning

./src/helper_functions/IncrementalDataset.py

Training with default setting

nohup python3 -u main.py > csc.log &

Citation

If you find our work useful for your research, please cite our work:

@inproceedings{du2024confidence,
  title={Confidence Self-Calibration for Multi-Label Class-Incremental Learning},
  author={Du, Kaile and Zhou, Yifan and Lyu, Fan and Li, Yuyang and Lu, Chen and Liu, Guangcan},
  booktitle={Proceedings of the European Conference on Computer Vision},
  pages={234--252},
  year={2024}
}