Research map / Detection and segmentation / Open-set and few-shot detection
Iterative Shrinking for Referring Expression Grounding Using Deep Reinforcement Learning
Mingjie Sun, Jimin Xiao, Eng Gee Lim
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 |
|---|---|---|---|
| use iterative shrinking with an RL agent instead of feature-point matching | proposal-free referring expression grounding · RefCOCOg test | accuracy: 65.73 -> 70.05 | Improved |
| use iterative shrinking with an RL agent instead of feature-point matching | proposal-free referring expression grounding · RefCOCO testA and testB | accuracy: 78.93 -> 74.27; 73.09 -> 68.10 | Mixed |
| use iterative shrinking with an RL agent | referring expression grounding · RefCOCOg, RefCOCO+, RefCOCO average | accuracy: 1.33% | Improved |
| add triad-based linguistic features instead of BERT features (ablation) | ISREG · RefCOCO testA | accuracy: 67.52 -> 74.27 | Improved |
| add triad-based linguistic features instead of BERT features (ablation) | ISREG · RefCOCOg test | accuracy: 60.17 -> 70.05 | Improved |
| add multi-scale visual features (ablation) | ISREG · RefCOCO testA | accuracy: 73.33 -> 74.27 | Improved |
| add multi-scale visual features (ablation) | ISREG · RefCOCOg test | accuracy: 66.35 -> 70.05 | Improved |
| add spatial features to the state (ablation) | ISREG · RefCOCO testA | accuracy: 62.42 -> 74.27 | Improved |
| add spatial features to the state (ablation) | ISREG · RefCOCOg test | accuracy: 54.88 -> 70.05 | Improved |
| use adaptive rather than fixed shrinking stride (ablation) | ISREG · RefCOCO testA and RefCOCOg test | accuracy: 45.70 -> 74.27; 39.46 -> 70.05 | Improved |
| add bounding-box refinement (ablation) | ISREG · RefCOCO testA and RefCOCOg test | accuracy: 70.18 -> 74.27; 65.62 -> 70.05 | Improved |
| use reinforcement learning rather than supervised training (ablation) | ISREG · RefCOCO testA and RefCOCOg test | accuracy: 63.84 -> 74.27; 61.36 -> 70.05 | Improved |
Closest papers by meaning
- Adaptive Reconstruction Network for Weakly Supervised Referring Expression Grounding (ICCV 2019 · similarity 0.89)
- Using Syntax to Ground Referring Expressions in Natural Images (AAAI 2018 · similarity 0.89)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.88)
- Query-Guided Regression Network With Context Policy for Phrase Grounding (ICCV 2017 · similarity 0.88)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.86)
- ScanFormer: Referring Expression Comprehension by Iteratively Scanning (CVPR 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.
