Research map / Vision-language and multimodal / Human and embodied modeling
Learning Structured Visual Compositional Representations for Weakly Supervised Referring Expression Comprehension
Lian Xu, Mohammed Bennamoun, Farid Boussaid, Hamid Laga, Yulan Guo, Dan Xu
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 Unary Object Representation (UOR) (ablation) | SVCR baseline (DINO+Depth) · RefCOCO testB | accuracy IoU@0.5 (%): 59.94 -> 64.30 | Improved |
| add Pairwise Relational Representation (PRR) (ablation) | SVCR baseline (DINO+Depth) · RefCOCO testB | accuracy IoU@0.5 (%): 59.94 -> 63.47 | Improved |
| add PRR (RefCOCOg, long expressions) (ablation) | SVCR baseline (DINO+Depth) · RefCOCOg val | accuracy IoU@0.5 (%): 51.08 -> 54.31 | Improved |
| add both UOR and PRR with compositional alignment (ablation) | SVCR baseline (DINO+Depth) · RefCOCOg val | accuracy IoU@0.5 (%): 51.08 -> 56.99 | Improved |
| add UOR and PRR to DINO+CLIP baseline (ablation) | SVCR baseline (DINO+CLIP) · RefCOCO testB | accuracy IoU@0.5 (%): 56.09 -> 60.43 | Improved |
| add UOR and PRR to WeakMCN (ViT-tiny-SAM) (ablation) | WeakMCN* · RefCOCO testB | accuracy IoU@0.5 (%): 62.00 -> 66.10 | Improved |
| add UOR and PRR to WeakMCN (ViT-small-SAM) (ablation) | WeakMCN dagger · RefCOCO testB | accuracy IoU@0.5 (%): 62.63 -> 64.89 | Improved |
| use explicit pairwise PRR instead of unary URR (ablation) | SVCR baseline (DINO+Depth) · RefCOCOg val | accuracy IoU@0.5 (%): 53.04 -> 54.31 | Improved |
| add diversity loss Ldiv (ablation) | SVCR · RefCOCO testB | accuracy IoU@0.5 (%): 64.59 -> 65.48 | Improved |
| add consistency loss Lcons (ablation) | SVCR · RefCOCO+ testB | accuracy IoU@0.5 (%): 43.58 -> 44.39 | Improved |
| add both Ldiv and Lcons (ablation) | SVCR · RefCOCO testB | accuracy IoU@0.5 (%): 64.59 -> 65.55 | Improved |
| increase anchors K from 17 to 34 (ablation) | SVCR · RefCOCOg val | accuracy IoU@0.5 (%): 56.99 -> 53.08 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DViN — DViN: Dynamic Visual Routing Network for Weakly Supervised Referring Expression Comprehension (CVPR 2025) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DViN: Dynamic Visual Routing Network for Weakly Supervised Referring Expression Comprehension (CVPR 2025 · similarity 0.91)
- Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks (ICLR 2026 · similarity 0.89)
- RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension (AAAI 2025 · similarity 0.88)
- Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models (CVPR 2025 · similarity 0.88)
- Zero-shot Referring Expression Comprehension via Structural Similarity Between Images and Captions (CVPR 2024 · similarity 0.88)
- Boosting Weakly Supervised Referring Image Segmentation via Progressive Comprehension (NeurIPS 2024 · 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.
