Research map / Detection and segmentation / Open-set and few-shot detection
Generalizable Object Re-Identification via Visual In-Context Prompting
Zhizhong Huang, Xiaoming Liu
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 ICL-generated visual prompts to triplet fine-tuned DINOv2 (ablation) | Triplet · ShopID10K (novel categories) | mAP: 50.3 -> 56.3 | Improved |
| add patch alignment loss (ablation) | VICP · ShopID10K (novel categories) | mAP: 56.3 -> 58.5 | Improved |
| add patch alignment loss (ablation) | VICP · PetFace (novel categories) | mAP: 50.8 -> 51.2 | Improved |
| use LLM in-context output directly instead of visual prompts (ablation) | VICP · ShopID10K (novel categories) | mAP: 56.3 -> 39.6 | Hurt |
| use 32 in-context examples instead of 64 (ablation) | VICP · ShopID10K (novel categories) | mAP: 58.5 -> 55.7 | Hurt |
| use 128 in-context examples instead of 64 (ablation) | VICP · ShopID10K (novel categories) | mAP: 58.5 -> 56.8 | Hurt |
| use 16 latent tokens instead of 32 (ablation) | VICP · ShopID10K (novel categories) | mAP: 58.5 -> 55.9 | Hurt |
| use 64 latent tokens instead of 32 (ablation) | VICP · ShopID10K (novel categories) | mAP: 58.5 -> 57.1 | Hurt |
| fine-tune DINOv2 with triplet loss on base categories | DINOv2 · ShopID10K (novel categories) | mAP: 34.1 -> 50.3 | Improved |
| use triplet loss instead of ArcFace | DINOv2 fine-tuning · ShopID10K (novel categories) | mAP: 45.3 -> 50.3 | Improved |
| add few-shot fine-tuning (Triplet+) | Triplet · ShopID10K (novel categories) | mAP: 50.3 -> 54.8 | Improved |
| replace Triplet+ with VICP prompts | Triplet+ · ShopID10K (novel categories) | mAP: 54.8 -> 58.5 | Improved |
Closest papers by meaning
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.87)
- Understanding Model Reprogramming for CLIP via Decoupling Visual Prompts (ICML 2025 · similarity 0.86)
- CLIP-driven View-aware Prompt Learning for Unsupervised Vehicle Re-identification (AAAI 2025 · similarity 0.85)
- A Pedestrian is Worth One Prompt: Towards Language Guidance Person Re-Identification (CVPR 2024 · similarity 0.85)
- TextRefiner: Internal Visual Feature as Efficient Refiner for Vision-Language Models Prompt Tuning (AAAI 2025 · similarity 0.85)
- MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-Identification (NeurIPS 2025 · 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.
