Research map / Detection and segmentation / Open-set and few-shot detection
CLIP-S4: Language-Guided Self-Supervised Semantic Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 pixel-segment contrastive learning to embedding and semantic consistency (ablation) | CLIP-S^4 · Pascal Context language-driven segmentation | mIoU: 31.3 -> 33.6 | Helped |
| Add embedding consistency to pixel-segment contrastive learning (ablation) | CLIP-S^4 · Pascal Context language-driven segmentation | mIoU: 32.9 -> 33.6 | Helped |
| Add semantic consistency to pixel-segment contrastive learning (ablation) | CLIP-S^4 · Pascal Context language-driven segmentation | mIoU: 24.3 -> 33.6 | Helped |
| Use embedding consistency without semantic consistency (ablation) | CLIP-S^4 · Pascal Context language-driven segmentation | mIoU: 33.6 -> 24.3 | Hurt |
| Use semantic consistency without embedding consistency (ablation) | CLIP-S^4 · Pascal Context language-driven segmentation | mIoU: 33.6 -> 32.9 | Hurt |
| Vary the number of unknown class prototypes from 16 to 256 (ablation) | CLIP-S^4 · Pascal VOC 2012 unsupervised segmentation | mIoU: 71.6; 71.9; 72.0; 72.2; 71.9 | No clear effect |
| Use CLIP-S^4 instead of MaskCLIP+ | Language-driven segmentation · Pascal Context | mIoU: +5.1 | Helped |
| Use CLIP-S^4 instead of MaskCLIP+ | Language-driven segmentation · COCO-Stuff | mIoU: +2.8 | Helped |
| Use CLIP-S^4 instead of MaskCLIP+ | Unsupervised semantic segmentation · Pascal VOC 2012 | mIoU (linear classifier): +3.0 | Helped |
| Use CLIP-S^4 instead of MaskCLIP+ | Class-free semantic segmentation · Pascal Context unknown classes, folds 0–3 | mIoU_u: +28.4; +19.3; +17.6; +24.1 | Mixed |
Closest papers by meaning
- Towards Open-Vocabulary Semantic Segmentation Without Semantic Labels (NeurIPS 2024 · similarity 0.89)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.89)
- CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.88)
- Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation (CVPR 2024 · similarity 0.88)
- Cascade-CLIP: Cascaded Vision-Language Embeddings Alignment for Zero-Shot Semantic Segmentation (ICML 2024 · similarity 0.87)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (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.