Research map / Detection and segmentation / Open-set and few-shot detection
Meta Compositional Referring Expression 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 |
|---|---|---|---|
| Apply MCRES meta-training to LAVT | LAVT · RefCOCO, RefCOCO+, and RefCOCOg test splits | overall IoU: RefCOCO val 72.73 -> 74.92; testA 75.82 -> 76.98; testB 68.79 -> 70.84; RefCOCO+ val 62.14 -> 64.32; testA 68.38 -> 69.6 | Helped |
| Apply MCRES meta-training | MCN, SeqTR, and LAVT · RefCOCO Novel subsets | overall IoU: MCN val 53.17 -> 57.38; testA 55.43 -> 59.66; testB 50.64 -> 55.87; SeqTR val 64.24 -> 67.31; testA 65.71 -> 69.48; test | Helped |
| Omit the word-word virtual testing set (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 73.43 -> 74.92 | Hurt |
| Omit the word-phrase virtual testing set (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 73.68 -> 74.92 | Hurt |
| Omit the phrase-phrase virtual testing set (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 73.59 -> 74.92 | Hurt |
| Use meta optimization instead of conventional non-meta training (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 72.76 -> 74.92 | Helped |
| Use separate virtual testing sets for each composition level instead of one combined set (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 73.64 -> 74.92 | Helped |
| Add curriculum learning over composition levels (ablation) | LAVT + MCRES with multiple virtual testing sets · RefCOCO val | overall IoU: 74.02 -> 74.92 | Helped |
| Change virtual training/testing data ratio to 50%:50% (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 74.78 vs. 74.92 | No clear effect |
| Change virtual training/testing data ratio to 70%:30% (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 74.74 vs. 74.92 | No clear effect |
| Select virtual testing samples randomly rather than for novel compositions (ablation) | LAVT + MCRES · RefCOCO val | overall IoU: 72.81 -> 74.92 | Hurt |
| Give the baseline additional gradient updates without MCRES (ablation) | LAVT baseline · RefCOCO val | overall IoU: 72.73 -> 72.74 | No clear effect |
Closest papers by meaning
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.88)
- Consistency of Compositional Generalization Across Multiple Levels (AAAI 2025 · similarity 0.87)
- RESAnything: Attribute Prompting for Arbitrary Referring Segmentation (NeurIPS 2025 · similarity 0.87)
- Rethinking Two-Stage Referring Expression Comprehension: A Novel Grounding and Segmentation Method Modulated by Point (AAAI 2024 · similarity 0.87)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.86)
- IteRPrimE: Zero-shot Referring Image Segmentation with Iterative Grad-CAM Refinement and Primary Word Emphasis (AAAI 2025 · similarity 0.85)
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.