Requirements

• Python3
• PyTorch (> 1.0)
• NumPy
• tqdm

Datasets

1. Download four public benchmarks for fine-grained dataset

   • CUB-200-2011
   • MIT-67: Indoor Scene Recognition
   • Stanford Dogs Dataset
   • FGVC-Aircraft Benchmark

2. Extract the tgz or zip file into ./data/ (Exceptionally, for CUB-200-2011, put the files in a ./data/CUB200)

Train Examples

python train.py \
--model resnet18 \
--dataset mit \
--alpha 32 \
--mrg 0.1 \
--lr 1e-4 \
--warm 5 \
--epochs 60 \
--batch-size 120 \

Acknowledgements

This repository is heavily built based on the following repository:

• Proxy Anchor-based Unsupervised Learning for Continuous Generalized Category Discovery

License

This project is licensed under the MIT License - see the LICENSE file for details.