Research map / Vision-language and multimodal / Cross-modal learning
Text-based Aerial-Ground Person Retrieval
Xinyu Zhou, Yu Wu, Jiayao Ma, Wenhao Wang, Min Cao, Mang Ye
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 HR-MoE blocks to image encoder (ablation) | TBPS-CLIP baseline · TAG-PEDES | R@1: 59.09 -> 61.60 | Improved |
| use vanilla MoE instead of ViT block (ablation) | TBPS-CLIP baseline · TAG-PEDES | R@1: 59.09 -> 60.26 | Improved |
| hierarchical routing and expert grouping vs vanilla MoE (ablation) | TAG-CLIP · TAG-PEDES | R@1: 60.26 -> 61.60 | Improved |
| add orthogonal loss on top of HR-MoE + view loss (ablation) | TAG-CLIP · TAG-PEDES | R@1: 61.55 -> 62.64 | Improved |
| add view loss on top of HR-MoE + orthogonal loss (ablation) | TAG-CLIP · TAG-PEDES | R@1: 62.32 -> 62.64 | No clear effect |
| add view and orthogonal losses to HR-MoE (ablation) | TAG-CLIP · TAG-PEDES | mAP: 48.94 -> 49.27 | No clear effect |
| train on aerial+ground instead of ground only (ablation) | TAG-CLIP · TAG-PEDES subset, ground test | R@1: 63.57 -> 64.49 | Improved |
| train on aerial+ground instead of aerial only (ablation) | TAG-CLIP · TAG-PEDES subset, aerial test | R@1: 54.04 -> 54.64 | Improved |
| train on aerial+ground instead of aerial only (ablation) | TAG-CLIP · TAG-PEDES subset, aerial test | R@5: 73.74 -> 73.01 | Hurt |
| TAG-CLIP vs RDE on TAG-PEDES | RDE · TAG-PEDES | R@1: 61.58 -> 62.64 | Improved |
| TAG-CLIP vs RDE on ground-only CUHK-PEDES | RDE · CUHK-PEDES | R@1: 75.94 -> 74.38 | Hurt |
| TAG-CLIP vs TBPS-CLIP on ground-only CUHK-PEDES | TBPS-CLIP · CUHK-PEDES | R@1: 72.66 -> 74.38 | Improved |
Closest papers by meaning
- Automatic Synthetic Data and Fine-grained Adaptive Feature Alignment for Composed Person Retrieval (NeurIPS 2025 · similarity 0.84)
- An Empirical Study of CLIP for Text-Based Person Search (AAAI 2024 · similarity 0.83)
- Adaptive Uncertainty-Based Learning for Text-Based Person Retrieval (AAAI 2024 · similarity 0.83)
- Exploring the Potential of Large Vision-Language Models for Unsupervised Text-Based Person Retrieval (AAAI 2025 · similarity 0.83)
- DM-Adapter: Domain-Aware Mixture-of-Adapters for Text-Based Person Retrieval (AAAI 2025 · similarity 0.83)
- Causality-Inspired Invariant Representation Learning for Text-Based Person Retrieval (AAAI 2024 · similarity 0.82)
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.
