Research map / Label-efficient and robust learning / Recognition and long-tail
DeepChange: A Long-Term Person Re-Identification Benchmark with Clothes Change
Peng Xu, Xiatian Zhu
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 deeper RGB backbones (ResNet152 instead of ResNet18) (ablation) | ResNet re-id · DeepChange test set | mAP (%): 08.44 -> 11.49 | Improved |
| use lightweight networks such as ShuffleNetv2 | re-id CNN · DeepChange test set | mAP (%): ShuffleNetv2_x0.5 04.00 vs ResNet50 09.62 | Hurt |
| feed grayscale or edge maps alone instead of RGB (ablation) | ResNet50 · DeepChange test set | mAP (%): 09.62 -> 06.96 (gray), 03.17 (edge) | Hurt |
| add grayscale or edge map branch via late fusion to ResNet50 (ablation) | ResNet50 · DeepChange test set | mAP (%): 09.62 -> 10.22 (R,G), 10.43 (R,E), 11.03 (R,G,E) | Improved |
| add body keypoint modality via late fusion to ResNet50 (ablation) | ResNet50 · DeepChange test set | mAP (%): 09.62 -> 09.54 | No clear effect |
| add RGB+grayscale multimodal fusion to state-of-the-art BNNeck re-id ResNet50 (ablation) | BNNeck re-id ResNet50 · DeepChange test set | mAP (%): 12.98 -> 13.12 | Improved |
| use early-fusion ViT with 0.3 weighted grayscale tokens added to RGB (ablation) | ViT B16 · DeepChange test set | mAP (%): 14.98 -> 15.19 | Improved |
| switch from CNN to transformer (ViT B16 vs ResNet50) | re-id backbone · DeepChange test set | mAP (%): 09.62 -> 14.98 | Improved |
| use ReIDCaps without auxiliary or capsule components (ablation) | ReIDCaps · DeepChange test set | mAP (%): 13.25 -> 09.25 (no auxiliary DenseNet121 variant) | Hurt |
Closest papers by meaning
- Person Re-Identification by Contour Sketch Under Moderate Clothing Change (TPAMI 2019 · similarity 0.90)
- Towards Realistic Lifelong Re-identification: Identity Recurrence with Changing Clothes (ICML 2026 · similarity 0.89)
- DIFFER: Disentangling Identity Features via Semantic Cues for Clothes-Changing Person Re-ID (CVPR 2025 · similarity 0.89)
- DKC: Differentiated Knowledge Consolidation for Cloth-Hybrid Lifelong Person Re-identification (CVPR 2025 · similarity 0.88)
- Identity-Clothing Similarity Modeling for Unsupervised Clothing Change Person Re-Identification (CVPR 2025 · similarity 0.87)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 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.
