Research map / Vision-language and multimodal / Human and embodied modeling
Consensus-Aware Visual-Semantic Embedding for Image-Text Matching
Haoran Wang, Ying Zhang, Zhong Ji, Yanwei Pang, Lin Ma
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 graph embedding (GCN over concept graph) (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 71.5 -> 74.8 | Improved |
| add confidence scaling function (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 74.5 -> 74.8 (text), 58.8 -> 59.9 (image) | Improved |
| add concept label prior for text consensus (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 72.5 -> 74.8 | Improved |
| add separate ranking losses on instance and consensus representations (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 72.9 -> 74.8 | Improved |
| add KL divergence between visual and textual concept scores (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 74.4 -> 74.8 | Improved |
| fuse consensus-level with instance-level representation instead of instance only (beta=1) (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 71.2 -> 74.8 | Improved |
| fuse instance-level with consensus-level instead of consensus only (beta=0) (ablation) | CVSE · MSCOCO 1K test | text retrieval R@1: 47.6 -> 74.8 | Improved |
| add consensus knowledge in MSCOCO-to-Flickr30k transfer | CVSE · Flickr30k (transfer from MSCOCO) | text retrieval R@1: 49.1 -> 56.4 | Improved |
| add consensus knowledge in MSCOCO-to-Flickr30k transfer | CVSE · Flickr30k (transfer from MSCOCO) | image retrieval R@1: 36.4 -> 39.9 | Improved |
Closest papers by meaning
- Image-Text Embedding Learning via Visual and Textual Semantic Reasoning (TPAMI 2022 · similarity 0.92)
- HAL: Improved Text-Image Matching by Mitigating Visual Semantic Hubs (AAAI 2020 · similarity 0.91)
- Asymmetric Visual Semantic Embedding Framework for Efficient Vision-Language Alignment (AAAI 2025 · similarity 0.90)
- Composing Object Relations and Attributes for Image-Text Matching (CVPR 2024 · similarity 0.89)
- Text as Partial Constraint: Core–Residual Alignment for Robust Vision–Language Learning (NeurIPS 2026 · similarity 0.88)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 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.
