Research map / Vision-language and multimodal / Cross-modal learning
Bi-Directional Relationship Inferring Network for Referring Image Segmentation
Check 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 bi-directional cross-modal attention and gated multi-level fusion (ablation) | DeepLab ResNet-101v2 referring segmentation network · UNC val | overall IoU: 52.26 -> 61.35 | Helped |
| Add vision-guided linguistic attention to the language-guided visual relationship module (ablation) | BCAM · UNC val | overall IoU: 55.14 -> 56.76 | Helped |
| Add language guidance to visual relationship modeling (ablation) | BCAM · UNC val | overall IoU: 53.91 -> 56.76 | Helped |
| Use gated top-down and bottom-up fusion rather than direct feature summation (ablation) | BRINet · UNC val | overall IoU: 59.59 -> 61.35 | Helped |
| Combine top-down and bottom-up fusion instead of using either direction alone (ablation) | BRINet · UNC val | overall IoU: 60.32 -> 61.35; 60.55 -> 61.35 | Helped |
| Use three feature scales rather than f4 + f5 (ablation) | BRINet · UNC val | overall IoU: 60.09 -> 61.35 | Helped |
| Use GBFM rather than ConvLSTM for multi-level feature fusion (ablation) | BRINet · UNC val | overall IoU: 58.78 -> 61.35 | Helped |
| Add spatial coordinate features (ablation) | BRINet · UNC val | overall IoU: 58.91 -> 61.35 | Helped |
| Add DenseCRF post-processing | BRINet · ReferIt test | overall IoU: 63.11 -> 63.46 | Helped |
| Add DenseCRF post-processing | BRINet · UNC testA | overall IoU: 62.99 -> 63.37 | No clear effect |
| Add DenseCRF post-processing | BRINet · UNC+ testB | overall IoU: 42.11 -> 42.13 | No clear effect |
| Use BRINet rather than STEP | Referring image segmentation · UNC testA | overall IoU: 63.46 -> 62.99 | No clear effect |
Closest papers by meaning
- X-RefSeg3D: Enhancing Referring 3D Instance Segmentation via Structured Cross-Modal Graph Neural Networks (AAAI 2024 · similarity 0.85)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.84)
- Keep the Balance: A Parameter-Efficient Symmetrical Framework for RGB+X Semantic Segmentation (CVPR 2025 · similarity 0.84)
- Remodeling Semantic Relationships in Vision-Language Fine-Tuning (AAAI 2026 · similarity 0.84)
- See Less, See Right: Bi-directional Perceptual Shaping For Multimodal Reasoning (CVPR 2026 · similarity 0.84)
- Perceiving Longer Sequences With Bi-Directional Cross-Attention Transformers (NeurIPS 2024 · similarity 0.84)
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.