lucasdavid / afhp

Adversarial Feature Hallucination in a Supervised Contrastive Space for Few-Shot Learning of Provenance in Paintings

Adversarial Feature Hallucination in a Supervised Contrastive Space for Few-Shot Learning of Provenance in Paintings

Official implementation of the paper Adversarial Feature Hallucination in a Supervised Contrastive Space for Few-Shot Learning of Provenance in Paintings, presented in LA-CCI 2023.

Summary

Adversarial Feature Hallucination Networks for Few-Shot Learning (AFHN) over a Supervised Contrastive space, applied to Few-Shot tasks sampled from the Painter by Numbers dataset.

Building and Running

Add the data to the ./data folder (or create a symbolic link with ln -s /path/to/dataset data).

You can run all configurations with:

$ ./runners/1-run-all.sh

To run the best configuration (our approach, last configuration in run-all.sh):

$ python run.py --data_split frequent --strategy supcon_mh --backbone_train_epochs 0

Citation

If our code be useful for you, please consider citing our paper using the following BibTeX entry.

@INPROCEEDINGS{10409405,
  author={David, Lucas and Liborio, Luis and Paiva, Guilherme and Severo, Marianna and Valle, Eduardo and Pedrini, Helio and Dias, Zanoni},
  booktitle={2023 IEEE Latin American Conference on Computational Intelligence (LA-CCI)},
  title={Adversarial Feature Hallucination in a Supervised Contrastive Space for Few-Shot Learning of Provenance in Paintings}, 
  year={2023},
  volume={},
  number={},
  pages={1-6},
  doi={10.1109/LA-CCI58595.2023.10409405}}