Research map / Vision-language and multimodal / Human and embodied modeling
Integrating Language Guidance Into Vision-Based Deep Metric Learning
Check 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 expert-label language guidance | Multisimilarity · CUB200-2011 | R@1: 62.8 ± 0.2 -> 67.3 ± 0.2 | Helped |
| Add top-5 ImageNet pseudolabel language guidance | Multisimilarity · CARS196 | R@1: 81.6 ± 0.3 -> 85.4 ± 0.2 | Helped |
| Add expert-label language guidance | Multisimilarity · SOP | R@1: 76.0 ± 0.1 -> 76.0 ± 0.2 | No clear effect |
| Add pseudolabel language guidance | Margin · SOP | R@1: 78.0 ± 0.3 -> 78.3 ± 0.2 | Helped |
| Add top-5 pseudolabel guidance | Multisimilarity + S2SD · CARS196 | R@1: 86.5% -> 88.4% | Helped |
| Use language-adjusted mining instead of similarity distillation (ablation) | Multisimilarity + ELG · CUB200-2011 | R@1: 67.3 ± 0.2 -> 64.1 ± 0.3 | Hurt |
| Use row-wise L2 rather than KL similarity-matrix matching (ablation) | Multisimilarity + ELG · CUB200-2011 | R@1: 67.3 ± 0.2 -> 65.2 ± 0.2 | Hurt |
| Allow gradients into the language model (ablation) | Multisimilarity + ELG · CUB200-2011 | R@1: 67.3 ± 0.2 -> 66.3 ± 0.4 | Hurt |
| Use CLIP-style direct image-language contrastive training (ablation) | Multisimilarity + language guidance · CUB200-2011 | R@1: 67.3% -> 63.7% | Hurt |
| Increase pseudolabel count from one to five (ablation) | PLG · CUB200-2011 | R@1: 66.2 ± 0.6 -> 67.1 ± 0.4 | Helped |
| Increase pseudolabel count from five to ten (ablation) | PLG · CUB200-2011 | R@1: 67.1 ± 0.4 -> 67.2 ± 0.2 | No clear effect |
| Use top-20 rather than top-5 pseudolabels (ablation) | PLG · CUB200-2011 | R@1: 67.1 ± 0.4 -> 67.0 ± 0.1 | Hurt |
Closest papers by meaning
- Learning Semantic Proxies from Visual Prompts for Parameter-Efficient Fine-Tuning in Deep Metric Learning (ICLR 2024 · similarity 0.89)
- Bridging Vision and Language Spaces with Assignment Prediction (ICLR 2024 · similarity 0.87)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · similarity 0.86)
- Potential Field Based Deep Metric Learning (CVPR 2025 · similarity 0.86)
- Deep Disentangled Metric Learning (AAAI 2025 · similarity 0.86)
- Language Model as Visual Explainer (NeurIPS 2024 · similarity 0.86)
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.