Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting
Abstract
Both generalized and incremental few-shot learning have to deal with three major challenges: learning novel classes from only few samples per class, preventing catastrophic forgetting of base classes, and classifier calibration across novel and base classes. In this work we propose a three-stage framework that allows to explicitly and effectively address these challenges. While the first phase learns base classes with many samples, the second phase learns a calibrated classifier for novel classes from few samples while also preventing catastrophic forgetting. In the final phase, calibration is achieved across all classes. We evaluate the proposed framework on four challenging benchmark datasets for image and video few-shot classification and obtain state-of-the-art results for both generalized and incremental few shot learning.
Cite
Text
Kukleva et al. "Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting." International Conference on Computer Vision, 2021. doi:10.1109/ICCV48922.2021.00889Markdown
[Kukleva et al. "Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting." International Conference on Computer Vision, 2021.](https://mlanthology.org/iccv/2021/kukleva2021iccv-generalized/) doi:10.1109/ICCV48922.2021.00889BibTeX
@inproceedings{kukleva2021iccv-generalized,
title = {{Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting}},
author = {Kukleva, Anna and Kuehne, Hilde and Schiele, Bernt},
booktitle = {International Conference on Computer Vision},
year = {2021},
pages = {9020-9029},
doi = {10.1109/ICCV48922.2021.00889},
url = {https://mlanthology.org/iccv/2021/kukleva2021iccv-generalized/}
}