English

Rethinking Prior Information Generation with CLIP for Few-Shot Segmentation

Computer Vision and Pattern Recognition 2024-05-15 v1

Abstract

Few-shot segmentation remains challenging due to the limitations of its labeling information for unseen classes. Most previous approaches rely on extracting high-level feature maps from the frozen visual encoder to compute the pixel-wise similarity as a key prior guidance for the decoder. However, such a prior representation suffers from coarse granularity and poor generalization to new classes since these high-level feature maps have obvious category bias. In this work, we propose to replace the visual prior representation with the visual-text alignment capacity to capture more reliable guidance and enhance the model generalization. Specifically, we design two kinds of training-free prior information generation strategy that attempts to utilize the semantic alignment capability of the Contrastive Language-Image Pre-training model (CLIP) to locate the target class. Besides, to acquire more accurate prior guidance, we build a high-order relationship of attention maps and utilize it to refine the initial prior information. Experiments on both the PASCAL-5{i} and COCO-20{i} datasets show that our method obtains a clearly substantial improvement and reaches the new state-of-the-art performance.

Keywords

Cite

@article{arxiv.2405.08458,
  title  = {Rethinking Prior Information Generation with CLIP for Few-Shot Segmentation},
  author = {Jin Wang and Bingfeng Zhang and Jian Pang and Honglong Chen and Weifeng Liu},
  journal= {arXiv preprint arXiv:2405.08458},
  year   = {2024}
}

Comments

Accepted by CVPR 2024; The camera-ready version

R2 v1 2026-06-28T16:26:40.313Z