Research map / Detection and segmentation / Open-set and few-shot detection
Locate Then Segment: A Strong Pipeline for Referring Image Segmentation
Ya Jing, Tao Kong, Wei Wang, Liang Wang, Lei Li, Tieniu Tan
Check 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 |
|---|---|---|---|
| Decouple referring segmentation into localization followed by mask refinement | LTS · RefCOCO+ and RefCOCOg | IoU: +3.2% on RefCOCO+ and +3.4% on RefCOCOg | Improved |
| Add cross-modal fusion (ablation) | LTS without fusion · RefCOCO val | IoU: 56.94 -> 65.05 | Improved |
| Add language-guided relevance filtering (ablation) | LTS without filter · RefCOCO val | IoU: 63.50 -> 65.05 | Improved |
| Add the segmentation refinement module (ablation) | LTS without segmentation · RefCOCO val | IoU: 56.94 -> 65.05 | Improved |
| Add locating loss (ablation) | LTS without locating loss · RefCOCO val | IoU: 65.05 -> 65.43 | Improved |
| Add LTS segmentation module to CMPC (ablation) | CMPC · RefCOCO val, testA, testB | IoU: 61.36 -> 62.75; 64.53 -> 65.34; 59.64 -> 61.08 | Improved |
| Replace relevance filtering with transformer localization (ablation) | LTS · RefCOCO, RefCOCO+, RefCOCOg val | IoU: 65.43 -> 66.15; 54.21 -> 54.52; 54.40 -> 54.51 | Improved |
| Apply relevance filtering twice instead of once (ablation) | LTS · RefCOCO val, testA, testB | IoU: 65.43 -> 66.04; 67.76 -> 68.68; 63.08 -> 63.27 | Improved |
| Apply relevance filtering three times instead of once (ablation) | LTS · RefCOCO val | IoU: 65.43 -> 65.54 | No clear effect |
| Generate the filtering kernel from each word rather than the sentence feature (ablation) | LTS · RefCOCO val, testA, testB | IoU: 65.43 -> 64.92; 67.76 -> 67.31; 63.08 -> 62.50 | Hurt |
| Increase input resolution (ablation) | LTS · RefCOCO val, testA, testB | IoU: 63.01 -> 65.90; 65.40 -> 68.16; 60.78 -> 63.45 | Improved |
| Use transformer, more filtering times, and larger input resolution (ablation) | LTS · RefCOCO, RefCOCO+, RefCOCOg val | IoU: 65.43 -> 66.75; 54.21 -> 54.94; 54.40 -> 54.51 | Improved |
Closest papers by meaning
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.91)
- Boosting Weakly Supervised Referring Image Segmentation via Progressive Comprehension (NeurIPS 2024 · similarity 0.91)
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.90)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.90)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- LESS: Label-Efficient and Single-Stage Referring 3D Segmentation (NeurIPS 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.
