Can Synthetic Images Conquer Forgetting? Beyond Unexplored Doubts in Few-Shot Class-Incremental Learning
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911063564353536 |
|---|---|
| author | Kim, Junsu Ku, Yunhoe Baek, Seungryul |
| author_facet | Kim, Junsu Ku, Yunhoe Baek, Seungryul |
| contents | Few-shot class-incremental learning (FSCIL) is challenging due to extremely limited training data; while aiming to reduce catastrophic forgetting and learn new information. We propose Diffusion-FSCIL, a novel approach that employs a text-to-image diffusion model as a frozen backbone. Our conjecture is that FSCIL can be tackled using a large generative model's capabilities benefiting from 1) generation ability via large-scale pre-training; 2) multi-scale representation; 3) representational flexibility through the text encoder. To maximize the representation capability, we propose to extract multiple complementary diffusion features to play roles as latent replay with slight support from feature distillation for preventing generative biases. Our framework realizes efficiency through 1) using a frozen backbone; 2) minimal trainable components; 3) batch processing of multiple feature extractions. Extensive experiments on CUB-200, \emph{mini}ImageNet, and CIFAR-100 show that Diffusion-FSCIL surpasses state-of-the-art methods, preserving performance on previously learned classes and adapting effectively to new ones. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_13739 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Can Synthetic Images Conquer Forgetting? Beyond Unexplored Doubts in Few-Shot Class-Incremental Learning Kim, Junsu Ku, Yunhoe Baek, Seungryul Computer Vision and Pattern Recognition Artificial Intelligence Few-shot class-incremental learning (FSCIL) is challenging due to extremely limited training data; while aiming to reduce catastrophic forgetting and learn new information. We propose Diffusion-FSCIL, a novel approach that employs a text-to-image diffusion model as a frozen backbone. Our conjecture is that FSCIL can be tackled using a large generative model's capabilities benefiting from 1) generation ability via large-scale pre-training; 2) multi-scale representation; 3) representational flexibility through the text encoder. To maximize the representation capability, we propose to extract multiple complementary diffusion features to play roles as latent replay with slight support from feature distillation for preventing generative biases. Our framework realizes efficiency through 1) using a frozen backbone; 2) minimal trainable components; 3) batch processing of multiple feature extractions. Extensive experiments on CUB-200, \emph{mini}ImageNet, and CIFAR-100 show that Diffusion-FSCIL surpasses state-of-the-art methods, preserving performance on previously learned classes and adapting effectively to new ones. |
| title | Can Synthetic Images Conquer Forgetting? Beyond Unexplored Doubts in Few-Shot Class-Incremental Learning |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2507.13739 |