A PyTorch Implementation of Adaptive Random Feature Regularization (CVPR'24)

[arXiv]

Requirements

Software Requirements

• CUDA >= 11.7

Python Requirements

• Please see pyproject.toml

Preparations

Here, we describe the preparation for the experiments on StanfordCars. You can use other datasets by modifying the preparation scripts.

Target Datset: StanfordCars

1. Download the dataset from here including {train,test}_annos.npz
2. Install the dataset into ./data/StanfordCars
3. Run the preparation script as follows:

cd ./data/StanfordCars/
python3 ../script/split_train_test.py

Example

Run Training of AdaRand with ResNet-50

python3 main/train.py --config_path config/04_adarand/rn50.yaml

Citation

@inproceedings{Yamaguchi_CVPR24_AdaRand,
  title={Adaptive Random Feature Regularization on Fine-tuning Deep Neural Networks},
  author={Yamaguchi, Shin'ya and Kanai, Sekitoshi and Adachi, Kazuki and Chijiwa, Daiki},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2024}
}