Research map / Detection and segmentation / Open-set and few-shot detection
Plug-in Feedback Self-adaptive Attention in CLIP for Training-free Open-Vocabulary Segmentation
Zhixiang Chi, Yanan Wu, Li Gu, Huan Liu, Ziqiang Wang, Yang Zhang, Yang Wang, Konstantinos Plataniotis
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add FSA to MaskCLIP | MaskCLIP (ViT-L/14) · 8 segmentation benchmarks | mIoU: 13.9 -> 32.6 | Improved |
| add FSA to SCLIP | SCLIP (ViT-L/14) · 8 segmentation benchmarks | mIoU: 29.0 -> 35.0 | Improved |
| add FSA to ClearCLIP | ClearCLIP (ViT-B/16) · 8 segmentation benchmarks | mIoU: 38.1 -> 38.8 | Improved |
| add FSA to ProxyCLIP | ProxyCLIP (ViT-H/14) · 8 segmentation benchmarks | mIoU: 44.4 -> 45.8 | Improved |
| add FSA to ProxyCLIP with MAE | ProxyCLIP (ViT-L/14) · 8 segmentation benchmarks | mIoU: 35.8 -> 38.3 | Improved |
| add FSA to Q-K attention | CLIP ViT-H/14 · 8 segmentation benchmarks | mIoU: 30.0 -> 34.6 | Improved |
| add confidence-based pruning instead of fixed ratio (ablation) | FSA on ProxyCLIP · 8 benchmarks | mIoU: 37.8 -> 39.4 | Improved |
| add adaptive exponential scaling (ablation) | FSA on ProxyCLIP · 8 benchmarks | mIoU: 39.4 -> 43.6 | Improved |
| add attention isolation (ablation) | FSA on MaskCLIP · 8 benchmarks | mIoU: 29.8 -> 32.6 | Improved |
| add attention isolation (ablation) | FSA on ProxyCLIP · 8 benchmarks | mIoU: 43.4 -> 43.6 | Improved |
| ensemble three dense outputs (ablation) | FSA on MaskCLIP · 8 benchmarks | mIoU: 32.1 -> 32.6 | Improved |
| use KL divergence instead of cosine similarity (ablation) | FSA on ProxyCLIP · 8 benchmarks | mIoU: 43.3 -> 43.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ClearCLIP — ClearCLIP: Decomposing CLIP Representations for Dense Vision-Language Inference (ECCV 2024) · better on 3 of 3 shared results
- ProxyCLIP — ProxyCLIP: Proxy Attention Improves CLIP for Open-Vocabulary Segmentation (ECCV 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ResCLIP: Residual Attention for Training-free Dense Vision-language Inference (CVPR 2025 · similarity 0.90)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.90)
- Diffusion Feedback Helps CLIP See Better (ICLR 2025 · similarity 0.89)
- Refining CLIP's Spatial Awareness: A Visual-Centric Perspective (ICLR 2025 · similarity 0.89)
- Understanding Fine-tuning CLIP for Open-vocabulary Semantic Segmentation in Hyperbolic Space (CVPR 2025 · similarity 0.89)
- What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update Perspective (NeurIPS 2026 · similarity 0.88)
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.
