Research map / Detection and segmentation / Open-set and few-shot detection
GRES: Generalized 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 |
|---|---|---|---|
| Add dynamic region feature aggregation instead of average-pooled region features (ablation) | ReLA · gRefCOCO val | gIoU: 60.93 -> 63.60; 2.67% gIoU gain | Helped |
| Add minimap supervision linking queries to image regions (ablation) | ReLA · gRefCOCO val | gIoU: 62.06 -> 63.60; 1.54% gIoU drop when removed | Helped |
| Use average-pooled image features rather than a rigid hard split before encoding (ablation) | ReLA · gRefCOCO val | gIoU: 55.34 -> 60.93; 5.59% gIoU gain | Helped |
| Add region-language cross-attention (ablation) | ReLA · gRefCOCO val | gIoU: 58.53 -> 60.53; 2% gIoU gain | Helped |
| Add region self-attention to model region-region dependencies (ablation) | ReLA · gRefCOCO val | gIoU: 60.53 -> 63.60; 3.85% gIoU gain | Helped |
| Add region-region and region-word relationship modeling together (ablation) | ReLA · gRefCOCO val | gIoU: +5.07% gIoU | Helped |
| Increase region grid from 4×4 to 8×8 to 10×10 (ablation) | ReLA · gRefCOCO val | gIoU: 57.01 -> 61.23 -> 63.60 | Helped |
| Increase region grid from 10×10 to 12×12 (ablation) | ReLA · gRefCOCO val | gIoU: 63.60 -> 63.50 | Hurt |
| Add ReLA relationship modeling to LAVT | LAVT · gRefCOCO val | cIoU: 57.64 -> 61.23 | Helped |
| Use a dedicated no-target classifier instead of detecting no-targets from the positive-pixel count | ReLA · gRefCOCO val | N-acc.: 49.96 -> 56.37 | Helped |
| Train on gRefCOCO rather than a single-target RES dataset (ablation) | RES models on multi-target and no-target expressions · gRefCOCO | segmentation quality | Helped |
Closest papers by meaning
- GSVA: Generalized Segmentation via Multimodal Large Language Models (CVPR 2024 · similarity 0.92)
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.92)
- RESAnything: Attribute Prompting for Arbitrary Referring Segmentation (NeurIPS 2025 · similarity 0.89)
- SAM as the Guide: Mastering Pseudo-Label Refinement in Semi-Supervised Referring Expression Segmentation (ICML 2024 · similarity 0.88)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.88)
- IteRPrimE: Zero-shot Referring Image Segmentation with Iterative Grad-CAM Refinement and Primary Word Emphasis (AAAI 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.