{"ID":2830218,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.10521","arxiv_id":"2512.10521","title":"Take a Peek: Efficient Encoder Adaptation for Few-Shot Semantic Segmentation via LoRA","abstract":"Few-shot semantic segmentation (FSS) aims to segment novel classes in query images using only a small annotated support set. While prior research has mainly focused on improving decoders, the encoder's limited ability to extract meaningful features for unseen classes remains a key bottleneck. In this work, we introduce \\textit{Take a Peek} (TaP), a simple yet effective method that enhances encoder adaptability for both FSS and cross-domain FSS (CD-FSS). TaP leverages Low-Rank Adaptation (LoRA) to fine-tune the encoder on the support set with minimal computational overhead, enabling fast adaptation to novel classes while mitigating catastrophic forgetting. Our method is model-agnostic and can be seamlessly integrated into existing FSS pipelines. Extensive experiments across multiple benchmarks--including COCO $20^i$, Pascal $5^i$, and cross-domain datasets such as DeepGlobe, ISIC, and Chest X-ray--demonstrate that TaP consistently improves segmentation performance across diverse models and shot settings. Notably, TaP delivers significant gains in complex multi-class scenarios, highlighting its practical effectiveness in realistic settings. A rank sensitivity analysis also shows that strong performance can be achieved even with low-rank adaptations, ensuring computational efficiency. By addressing a critical limitation in FSS--the encoder's generalization to novel classes--TaP paves the way toward more robust, efficient, and generalizable segmentation systems. The code is available at https://github.com/pasqualedem/TakeAPeek.","short_abstract":"Few-shot semantic segmentation (FSS) aims to segment novel classes in query images using only a small annotated support set. While prior research has mainly focused on improving decoders, the encoder's limited ability to extract meaningful features for unseen classes remains a key bottleneck. In this work, we introduce...","url_abs":"https://arxiv.org/abs/2512.10521","url_pdf":"https://arxiv.org/pdf/2512.10521v1","authors":"[\"Pasquale De Marinis\",\"Gennaro Vessio\",\"Giovanna Castellano\"]","published":"2025-12-11T10:47:01Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"LoRA\"]","has_code":false,"code_links":[{"ID":606010,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2830218,"paper_url":"https://arxiv.org/abs/2512.10521","paper_title":"Take a Peek: Efficient Encoder Adaptation for Few-Shot Semantic Segmentation via LoRA","repo_url":"https://github.com/pasqualedem/TakeAPeek","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}