Research map / Detection and segmentation / Open-set and few-shot detection
CoupAlign: Coupling Word-Pixel with Sentence-Mask Alignments for Referring Image Segmentation
Zicheng Zhang, Yi Zhu, Jianzhuang Liu, Xiaodan Liang, Wei Ke
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 |
|---|---|---|---|
| couple word-pixel alignment with sentence-mask alignment | CoupAlign · RefCOCO val | oIoU: 72.73 -> 74.70 | Improved |
| add bidirectional WPA (ablation) | CoupAlign · RefCOCO val | oIoU: 70.43 -> 74.70 | Improved |
| use bidirectional rather than unidirectional language-to-image WPA (ablation) | CoupAlign · RefCOCO val | oIoU: 72.70 -> 74.70 | Improved |
| add SMA mask constraints (ablation) | CoupAlign · RefCOCO val | oIoU: 73.02 -> 74.70 | Improved |
| add the auxiliary pixel-level contrastive loss (ablation) | CoupAlign · RefCOCO val | oIoU: 73.70 -> 74.70 | Improved |
| place WPA at stages 1-4 instead of stages 1-2 (ablation) | CoupAlign · RefCOCO val | oIoU: 72.74 -> 74.70 | Improved |
| place WPA at stages 1-4 instead of stages 3-4 (ablation) | CoupAlign · RefCOCO val | oIoU: 73.93 -> 74.70 | Improved |
| use WPA at stage 4 rather than stage 3 (ablation) | CoupAlign · RefCOCO val | oIoU: 72.53 -> 73.61 | Improved |
| increase mask-query count from 20 to 100 (ablation) | CoupAlign · RefCOCO val | mIoU: 74.77 -> 75.49 | Improved |
| increase mask-query count from 50 to 100 (ablation) | CoupAlign · RefCOCO val | prec@0.9: 31.88 -> 32.40 | Improved |
| apply CoupAlign instead of LAVT | Swin-B referring image segmentation · RefCOCO+ test A | oIoU: 68.38 -> 68.34 | No clear effect |
Closest papers by meaning
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.89)
- GOAL: Global-local Object Alignment Learning (CVPR 2025 · similarity 0.88)
- Pixel-Aligned Language Model (CVPR 2024 · similarity 0.88)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · similarity 0.88)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.87)
- DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment (CVPR 2025 · 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.
