Research map / Vision-language and multimodal / Human and embodied modeling
Divide and Align: Disentangled Vision-Language Learning for Text-Based Person Anomaly Search
Alex Ergasti, Tomaso Fontanini, Claudio Ferrari, Massimo Bertozzi, Andrea Prati
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 |
|---|---|---|---|
| use sigmoid weights instead of softmax in STP (ablation) | STP · PAB (0.1M) | R@1: 83.82 -> 84.78 | Improved |
| use STP instead of a cross-attention module (ablation) | STP · PAB (0.1M) | R@1: 83.92 -> 84.78 | Improved |
| add Feature Decoupling Loss (ablation) | STP · PAB (0.1M) | R@1: 83.15 -> 84.78 | Improved |
| add Feature Decoupling Loss (ablation) | STP · PAB (0.1M) | mAP: 90.52 -> 91.54 | Improved |
| add semantic-aware ITM loss (ablation) | STP+FDL · PAB (0.1M) | R@1: 84.78 -> 85.14 | Improved |
| add semantic-aware ITM loss (ablation) | STP+FDL · PAB (0.1M) | mAP: 91.54 -> 91.83 | Improved |
| combine appearance, action and background instead of appearance only (ablation) | STP · PAB (0.1M) | R@1: 83.67 -> 85.14 | Improved |
| combine all three components instead of action+background (ablation) | STP · PAB (0.1M) | R@1: 84.88 -> 85.14 | Improved |
| train on 1M instead of 0.1M samples (ablation) | SeDA · PAB | R@1: 85.14 -> 86.45 | Improved |
| use SeDA instead of CMP under multi-weather (ablation) | SeDA · PAB multi-weather (mean) | R@1: 67.12 -> 69.55 | Improved |
| use SeDA instead of CMP out-of-distribution (ablation) | SeDA · UCC OOD | R@1: 55.23 -> 58.97 | Improved |
| use SeDA instead of CMP on identity search | SeDA · PAB identity search | mAP: 88.01 -> 88.08 | No clear effect |
Closest papers by meaning
- Visual Perturbation for Text-Based Person Search (AAAI 2025 · similarity 0.88)
- Adaptive Uncertainty-Based Learning for Text-Based Person Retrieval (AAAI 2024 · similarity 0.87)
- Chat-based Person Retrieval via Dialogue-Refined Cross-Modal Alignment (CVPR 2025 · similarity 0.86)
- Human-centered Interactive Learning via MLLMs for Text-to-Image Person Re-identification (CVPR 2025 · similarity 0.86)
- Causality-Inspired Invariant Representation Learning for Text-Based Person Retrieval (AAAI 2024 · similarity 0.86)
- An Empirical Study of CLIP for Text-Based Person Search (AAAI 2024 · similarity 0.85)
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.
