Research map / Detection and segmentation / Open-set and few-shot detection
CP-CLIP: Customized Parameter Generation for Open-vocabulary Semantic Segmentation
Zelin Peng, Zhengqin Xu, Feilong Tang, Wei Shen
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 orthogonal adaptation (fixed Gaussian beta) to CP-CLIP (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 11.5 -> 12.8 | Improved |
| add customized parameter generator conditioned on local visual features (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 11.9 -> 12.8 | Improved |
| add customized parameter generator conditioned on local visual features (ablation) | CP-CLIP (ViT-B/16) · A-150 | mIoU: 31.9 -> 32.5 | Improved |
| freeze beta instead of training it (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 11.1 -> 12.8 | Improved |
| also fine-tune the image encoder (vs text only) (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 11.2 -> 12.8 | Improved |
| fine-tune CLIP text encoder with generated parameters (vs image only) (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 10.4 -> 12.8 | Improved |
| use CP-CLIP instead of plain LoRA (ablation) | LoRA (ViT-B/16) · A-847 | mIoU: 11.4 -> 12.8 | Improved |
| increase generator layers M from 3 to 4 (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 12.8 -> 12.7 | No clear effect |
| increase rank r from 32 to 64 (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 12.8 -> 12.6 | No clear effect |
| use high-level (layer 12) instead of layer 1 visual features as condition (ablation) | CP-CLIP (ViT-B/16) · A-847 | mIoU: 11.6 -> 12.8 | Improved |
| replace CAT-Seg with CP-CLIP | CAT-Seg (ViT-B/16) · A-847 | mIoU: 12.0 -> 12.8 | Improved |
| replace CAT-Seg with CP-CLIP | CAT-Seg (ViT-B/16) · PAS-20b | mIoU: 77.3 -> 77.1 | Hurt |
Closest papers by meaning
- Towards Open-Vocabulary Semantic Segmentation Without Semantic Labels (NeurIPS 2024 · similarity 0.90)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.90)
- Dual Semantic Guidance for Open Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.90)
- CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.90)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.90)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
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.
