Research map / Detection and segmentation / Open-set and few-shot detection
Class Distribution-induced Attention Map for Open-vocabulary Semantic Segmentations
Dong Un Kang, Hayeon Kim, Se Young Chun
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a class-distribution-induced attention map from CLIP for training-free open-vocabulary semantic segmentation to better localize class-related objects.
Problem
Open-vocabulary semantic segmentation with vision-language models struggles to accurately localize class-related objects, with CLIP-based methods producing patch-wise noisy class predictions.
Approach
Argues that CLIP-based prior works yield noisy patch-wise class predictions despite highly class-correlated attention, and derives a class-distribution-induced attention map to improve localization of class objects for zero-shot segmentation.
Surprising findings
- CLIP-based prior works produce patch-wise noisy class predictions even though their attention is highly class-correlated (from abstract)
Closest papers by meaning
- Leveraging Class Distributions in CLIP for Weakly Supervised Semantic Segmentation (CVPR 2026 · similarity 0.91)
- OPMapper: Enhancing Open-Vocabulary Semantic Segmentation with Multi-Guidance Information (NeurIPS 2025 · similarity 0.90)
- Novel Class Discovery in Semantic Segmentation (CVPR 2022 · similarity 0.89)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.87)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.87)
- CSL: Class-Agnostic Structure-Constrained Learning for Segmentation Including the Unseen (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
