Research map / Vision-language and multimodal / Human and embodied modeling
Expressing Objects Just Like Words: Recurrent Visual Embedding for Image-Text Matching
Read the paperCheck 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 recurrent visual embedding with text-ordered objects (ablation) | SCAN · MS-COCO 1K | sentence retrieval R@1: 72.7 -> 73.8 | Helped |
| add recurrent visual embedding with text-ordered objects (ablation) | SCAN · MS-COCO 1K | image retrieval R@1: 58.8 -> 60.5 | Helped |
| add recurrent visual embedding (ablation) | Multi-ATT · MS-COCO 1K | sentence retrieval R@1: 73.4 -> 75.3 | Helped |
| add recurrent visual embedding (ablation) | Multi-ATT · MS-COCO 1K | image retrieval R@1: 60.2 -> 62.5 | Helped |
| order objects by related word position rather than random order (ablation) | DP-RNN (SCAN-based) · MS-COCO 1K | image retrieval R@1: 59.0 -> 60.5 | Helped |
| order objects by related word position rather than random order (ablation) | DP-RNN (SCAN-based) · MS-COCO 1K | sentence retrieval R@1: 72.4 -> 73.8 | Helped |
| use Multi-attention Cross Matching instead of SCAN cross matching (both without RVE) (ablation) | Multi-ATT vs SCAN · MS-COCO 1K | sentence retrieval R@1: 72.7 -> 73.4 | Helped |
| increase early-selection d from 5 to 10 (ablation) | DP-RNN · MS-COCO 1K | image retrieval R@1: 61.6 -> 62.5 | Helped |
| increase early-selection d from 10 to 20 (ablation) | DP-RNN · MS-COCO 1K | image retrieval R@1: 62.5 -> 62.4 | No clear effect |
| increase early-selection d from 10 to 20 (ablation) | DP-RNN · MS-COCO 1K | per batch training time: 1.326 -> 1.482 | Hurt |
Closest papers by meaning
- Composing Object Relations and Attributes for Image-Text Matching (CVPR 2024 · similarity 0.92)
- End-to-End Convolutional Semantic Embeddings (CVPR 2018 · similarity 0.89)
- Training and Evaluating Multimodal Word Embeddings with Large-scale Web Annotated Images (NeurIPS 2016 · similarity 0.89)
- Learning Deep Structure-Preserving Image-Text Embeddings (CVPR 2016 · similarity 0.89)
- Instance-Aware Image and Sentence Matching With Selective Multimodal LSTM (CVPR 2017 · similarity 0.88)
- Text Embedding is Not All You Need: Attention Control for Text-to-Image Semantic Alignment with Text Self-Attention Maps (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.