Research map / Detection and segmentation / Open-set and few-shot detection
A Mutual Supervision Framework for Referring Expression Segmentation and Generation
Shijia Huang, Feng Li, Hao Zhang, Shilong Liu, Lei Zhang, Liwei Wang
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 end-to-end optimization of the proposal extractor (ablation) | REG model · RefCOCO+ testB | CIDEr: 0.791 -> 0.847 | Improved |
| add joint training with RES (ablation) | REG model · RefCOCO+ testB | CIDEr: 0.847 -> 0.879 | Improved |
| add the indicator module (ablation) | REG model · RefCOCO+ testB | CIDEr: 0.879 -> 0.895 | Improved |
| add disambiguation supervision (ablation) | REG model · RefCOCO+ testB | CIDEr: 0.895 -> 0.927 | Improved |
| add disambiguation supervision (ablation) | REG model · RefCOCO+ testB | Refer-back accuracy: 75.6% -> 79.1% | Improved |
| add CIDEr reward to unambiguity reward (ablation) | REG model · RefCOCO+ testB | CIDEr: 0.844 -> 0.927 | Improved |
| use softmax instead of sigmoid unambiguity reward (ablation) | REG model · RefCOCO+ testB | Refer-back accuracy: 74.5% -> 78.0% | Improved |
| add temperature tau smoothing to softmax reward (ablation) | REG model · RefCOCO+ testB | Refer-back accuracy: 78.0% -> 79.1% | Improved |
| add end-to-end optimization with RES loss (ablation) | RES model · RefCOCO val | oIoU: 68.98 -> 74.21 | Improved |
| add joint training with REG (ablation) | RES model · RefCOCO+ val | oIoU: 64.99 -> 66.21 | Improved |
| add generation supervision (ablation) | RES model · RefCOCO+ val | oIoU: 64.99 -> 67.24 | Improved |
| add generation supervision with all unfiltered pseudo-annotations without filtering/reweighting (ablation) | RES model · RefCOCO+ | mIoU | Hurt |
Closest papers by meaning
- Rethinking Two-Stage Referring Expression Comprehension: A Novel Grounding and Segmentation Method Modulated by Point (AAAI 2024 · similarity 0.87)
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.87)
- RESAnything: Attribute Prompting for Arbitrary Referring Segmentation (NeurIPS 2025 · similarity 0.86)
- 3D-STMN: Dependency-Driven Superpoint-Text Matching Network for End-to-End 3D Referring Expression Segmentation (AAAI 2024 · similarity 0.86)
- Position Is All You Need: A Free Lunch Token Compression Strategy for MLLM-based Referring Expression Segmentation (ICML 2026 · similarity 0.86)
- SAM as the Guide: Mastering Pseudo-Label Refinement in Semi-Supervised Referring Expression Segmentation (ICML 2024 · similarity 0.86)
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.
