gabmoreira / hyper

Code for Hyperbolic vs Euclidean Embeddings in Few-Shot Learning: Two Sides of the Same Coin, WACV (2024)

Training

Create an experiments folder in the parent directory of ./code/

mkdir ./experiments/

Edit the cfg dictionary in train.py according to the desired specifications Move to the parent directory and run the train.py script

python ./code/train.py

During training an automatic experiment name is generated and the best model weights will be stored at ./experiments/<experiment_name>/best_weights.pt

The cfg dictionary is saved as ./experiments/<experiment_name>/cfg.pt

The tracker CSV with train and validation accuracies and losses is saved as ./experiments/<experiment_name>/tracker.csv

Testing

For testing a trained model simply provide the experiment name and the shot/way/query regime you want to test it on

python ./code/test.py ./experiments/<experiment_name> <shot> <way> <query>

References

G Moreira, M Marques, JP Costeira, and A Hauptmann. "Hyperbolic vs Euclidean Embeddings in Few-Shot Learning: Two Sides of the Same Coin." WACV 2024 (To appear) arXiv:2309.10013.