Research map / Detection and segmentation / CNN features
ReCLIP++: Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation
Jingyun Wang, Guoliang Kang
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 decoder (ablation) | ReCLIP (rec.) · PASCAL VOC | mIoU: 58.5 -> 75.6 | Improved |
| add label hypothesis (ablation) | ReCLIP (rec.) + Decoder · PASCAL VOC | mIoU: 75.6 -> 77.3 | Improved |
| add independent bias encoding (addition fusion) (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 77.3 -> 81.1 | Improved |
| fuse W_r and W_p by matrix multiplication instead of addition (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 81.1 -> 85.4 | Improved |
| fix Query instead of making it learnable (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 11.6 -> 85.4 | Improved |
| use element-wise subtraction instead of addition (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 77.7 -> 85.4 | Improved |
| freeze CLIP instead of fine-tuning its image encoder (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 76.5 -> 85.4 | Improved |
| rectify both class and space preference instead of class only (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 76.4 -> 85.4 | Improved |
| rectify both class and space preference instead of space only (ablation) | ReCLIP++ · PASCAL Context | mIoU: 32.4 -> 36.1 | Improved |
| add Z as decoder input (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 73.6 -> 85.4 | Improved |
| add distillation stage (ablation) | ReCLIP++ · PASCAL VOC | mIoU: 85.4 -> 80.0 | Hurt |
| add distillation stage (ablation) | ReCLIP++ · ADE20K | mIoU: 16.4 -> 15.8 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ClearCLIP — ClearCLIP: Decomposing CLIP Representations for Dense Vision-Language Inference (ECCV 2024) · better on 5 of 5 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
- Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation (CVPR 2024 · similarity 0.98)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Understanding Fine-tuning CLIP for Open-vocabulary Semantic Segmentation in Hyperbolic Space (CVPR 2025 · similarity 0.87)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.87)
- Frozen CLIP: A Strong Backbone for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- un$^2$CLIP: Improving CLIP's Visual Detail Capturing Ability via Inverting unCLIP (NeurIPS 2025 · similarity 0.86)
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.
