We use the same backbone network and training strategies as 'S2M2_R'. Please refer to https://github.com/nupurkmr9/S2M2_fewshot for the backbone training.
After training the backbone as 'S2M2_R', extract features as below:
-
Create an empty 'checkpoints' directory.
-
Run:
python save_plk.py --dataset [miniImagenet/CUB]
https://drive.google.com/drive/folders/1IjqOYLRH0OwkMZo8Tp4EG02ltDppi61n?usp=sharing
After downloading the extracted features, please adjust your file path according to the code.
To evaluate our P3DC-Shot method, run:
python P3DC.py for 1-shot
python P3DC_5shot.py for 5-shot
Charting the Right Manifold: Manifold Mixup for Few-shot Learning
https://github.com/nupurkmr9/S2M2_fewshot
Free Lunch for Few-shot Learning: Distribution Calibration
https://github.com/ShuoYang-1998/Few_Shot_Distribution_Calibration