Research map / Vision-language and multimodal / Cross-modal learning
CAMP: Cross-Modal Adaptive Message Passing for Text-Image Retrieval
Zihao Wang, Xihui Liu, Hongsheng Li, Lu Sheng, Junjie Yan, Xiaogang Wang, Jing Shao
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 cross-modal message passing (vs base model without interaction) (ablation) | Base model · Flickr30K 1K test | caption retrieval R@1: 63.5 -> 68.1 | Improved |
| add cross-modal attention for message aggregation (vs averaging) (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 59.7 -> 68.1 | Improved |
| add cross-modal gated fusion on top of cross-modal attention (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 65.6 -> 68.1 | Improved |
| add adaptive gates to fusion (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 61.7 -> 68.1 | Improved |
| add residual connection to fusion (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 56.7 -> 68.1 | Improved |
| add attention-based fused feature aggregation (vs average pooling) (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 63.4 -> 68.1 | Improved |
| use element-wise add instead of concatenation fusion (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 66.3 -> 68.1 | Improved |
| use element-wise add instead of product fusion (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 61.5 -> 68.1 | Improved |
| use fused-feature MLP with BCE loss instead of joint embedding with ranking loss (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 62.0 -> 68.1 | Improved |
| use BCE loss instead of ranking loss on MLP score (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 60.9 -> 68.1 | Improved |
| add hardest negatives to BCE loss (ablation) | CAMP · Flickr30K 1K test | caption retrieval R@1: 65.5 -> 68.1 | Improved |
Later papers that compared with this method
2 later accepted papers report CAMP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Context-Aware Attention Network for Image-Text Retrieval (CVPR 2020 · better on 3 of 3 shared results)
- Multi-Modality Cross Attention Network for Image and Sentence Matching (CVPR 2020 · better on 1 of 2 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
- Look, Imagine and Match: Improving Textual-Visual Cross-Modal Retrieval With Generative Models (CVPR 2018 · similarity 0.90)
- Adaptive Cross-Modal Prototypes for Cross-Domain Visual-Language Retrieval (CVPR 2021 · similarity 0.90)
- Probabilistic Embeddings for Cross-Modal Retrieval (CVPR 2021 · similarity 0.90)
- Adaptive Prompt-Based Semantic Embedding with Inspire Potential of Implicit Knowledge for Cross-Modal Retrieval (AAAI 2025 · similarity 0.88)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 2025 · similarity 0.87)
- Cross-Modal and Uni-Modal Soft-Label Alignment for Image-Text Retrieval (AAAI 2024 · 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.
