Research map / Label-efficient and robust learning / Recognition and long-tail
UnrealPerson: An Adaptive Pipeline Towards Costless Person Re-Identification
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 |
|---|---|---|---|
| Train on Unreal synthetic data instead of RandPerson | Direct-transfer ReID model · MSMT17 | rank-1: 20.0 -> 38.5 | Helped |
| Pre-train on Unreal rather than ImageNet before Market fine-tuning | ReID model fine-tuned on Market · Market | rank-1: 91.4 -> 94.0 | Helped |
| Pre-train on Unreal rather than ImageNet before Duke fine-tuning | ReID model fine-tuned on Duke · Duke | rank-1: 82.1 -> 86.8 | Helped |
| Use Unreal as the labeled source for JVTC+ instead of the strongest listed source | JVTC+ · Duke | rank-1: 85.2 -> 88.3 | Helped |
| Use Unreal as the labeled source for JVTC+ instead of the strongest listed source | JVTC+ · MSMT17 | rank-1: 66.8 -> 68.2 | Helped |
| Use real clothing textures instead of generated regular textures (ablation) | Unreal synthetic ReID training data · Market | rank-1: 61.0 -> 64.5 | Helped |
| Add accessories (ablation) | Unreal synthetic ReID training data with real clothing textures · Duke | rank-1: 54.3 -> 57.0 | Helped |
| Add similar-appearance hard samples (ablation) | Unreal data with 800 IDs and 6 cameras · MSMT17 | rank-1: 21.6 -> 21.9 | Helped |
| Increase the number of cameras from 6 to 34 (ablation) | Unreal data with 800 identities · MSMT17 | rank-1: 18.5 -> 35.4 | Helped |
| Increase synthetic identities from 800 to 3,000 (ablation) | Unreal data with 34 cameras · MSMT17 | rank-1: 35.4 -> 38.5 | Helped |
| Use hard samples in a small 800-ID, 6-camera dataset (ablation) | Unreal synthetic ReID training data · Market | rank-1: 65.3 -> 65.2 | Hurt |
| Use hard samples in a small 800-ID, 6-camera dataset (ablation) | Unreal synthetic ReID training data · Duke | rank-1: 57.0 -> 56.7 | Hurt |
Closest papers by meaning
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.88)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.88)
- Person De-reidentification: A Variation-guided Identity Shift Modeling (CVPR 2025 · similarity 0.88)
- Instruct-ReID: A Multi-purpose Person Re-identification Task with Instructions (CVPR 2024 · similarity 0.87)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.87)
- Diffusion-based Synthetic Data Generation for Visible-Infrared Person Re-Identification (AAAI 2025 · 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.