Research map / Detection and segmentation / Weakly supervised detection and segmentation
Relation-aware Instance Refinement for Weakly Supervised Visual Grounding
Yongfei Liu, Bo Wan, Lin Ma, Xuming He
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 relation-aware instance refinement rather than Contrastive Learning | Weakly supervised visual grounding · Flickr30K Entities | Acc: 51.67 -> 59.27 | Improved |
| Use relation-aware instance refinement rather than Contrastive Learning | Weakly supervised visual grounding · Flickr30K Entities | PointIt: 76.74 -> 78.60 | Improved |
| Use relation-aware instance refinement rather than KPRN | Weakly supervised visual grounding · ReferItGame | Acc: 33.87 -> 37.68 | Improved |
| Use relation-aware instance refinement rather than KAC Net* | Weakly supervised visual grounding · ReferItGame | PointIt: 56.57 -> 58.96 | Improved |
| Add two-stage denoising (ablation) | Weakly supervised visual grounding baseline · Flickr30K Entities val | Acc: 48.18 -> 50.80 | Improved |
| Add self-taught regression with phrase features (ablation) | TSD grounding model · Flickr30K Entities val | Acc: 50.80 -> 56.88 | Improved |
| Include the noun phrase feature in proposal-offset estimation (ablation) | TSD + STR grounding model · Flickr30K Entities val | Acc: 54.05 -> 56.88 | Improved |
| Add visual object graph network and relation constraint (ablation) | TSD + STR grounding model · Flickr30K Entities val | Acc: 56.88 -> 58.30 | Improved |
| Use learned graph attention instead of uniform graph weights (ablation) | TSD + STR + VOGN + RC model · Flickr30K Entities val | Acc: 55.60 -> 58.30 | Improved |
| Remove self-taught regression loss (ablation) | Relation-aware Instance Refinement · Flickr30K Entities val | Acc | Hurt |
| Set retained proposals per phrase K to 3 instead of 5 (ablation) | Relation-aware Instance Refinement · Flickr30K Entities val | Acc: 58.30 -> 57.47 | Hurt |
| Set retained proposals per phrase K to 10 instead of 5 (ablation) | Relation-aware Instance Refinement · Flickr30K Entities val | Acc: 58.30 -> 57.15 | Hurt |
Closest papers by meaning
- Distilling Coarse-to-Fine Semantic Matching Knowledge for Weakly Supervised 3D Visual Grounding (ICCV 2023 · similarity 0.91)
- Learning Cross-Modal Context Graph for Visual Grounding (AAAI 2020 · similarity 0.91)
- Weakly-Supervised Generation and Grounding of Visual Descriptions With Conditional Generative Models (CVPR 2022 · similarity 0.90)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.90)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.90)
- AugRefer: Advancing 3D Visual Grounding via Cross-Modal Augmentation and Spatial Relation-based Referring (AAAI 2025 · similarity 0.88)
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.
