Research map / Label-efficient and robust learning / Recognition and long-tail
Temporal Model Adaptation for Person Re-identification
Niki Martinel, Abir Das, Christian Micheloni, Amit K. Roy-Chowdhury
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 incremental model updates (TMA1 to TMA4) | TMA · VIPeR | rank-1 accuracy (%): 35.13 -> 41.46 | Improved |
| add incremental model updates (TMA1 to TMA4) | TMA · PRID450S | rank-1 accuracy (%): 45.33 -> 52.89 | Improved |
| add incremental model updates (TMA1 to TMA4) | TMA with BoW · Market-1501 | rank-1 accuracy (%): 28.77 -> 44.74 | Improved |
| train on all data in one batch (TMA0) instead of 4 incremental updates (ablation) | TMA · VIPeR | rank-1 accuracy (%): 41.46 -> 43.83 | Mixed |
| fuse TMA0 scores with LADF scores | TMA0 · VIPeR | rank-1 accuracy (%): 43.83 -> 48.19 | Improved |
| replace BoW features with LOMO features (ablation) | TMA4 · Market-1501 | rank-1 accuracy (%): 44.74 -> 47.92 | Improved |
| add similarity measure to dissimilarity-only learning (ablation) | TMA · VIPeR | rank-1 accuracy (%): 7% at rank 1 | Improved |
| add l2,1 norm regularizers on projections (ablation) | TMA · VIPeR | rank-1 accuracy (%): degradation of 3% when removed | Improved |
| probe relevant set threshold epsilon 0.01 instead of 0.1 (ablation) | TMA4 · VIPeR | rank-1 accuracy (%): 41.46 -> 37.97 | Hurt |
| probe relevant set threshold epsilon 0.5 instead of 0.1 (ablation) | TMA4 · VIPeR | rank-1 accuracy (%): 41.46 -> 38.61 | Hurt |
| use graph-based relevant set selection instead of full supervised labeling (ablation) | TMA4 · VIPeR | rank-1 accuracy (%): 39.87 -> 41.46 | Improved |
| use unsupervised or semi-supervised pair labeling instead of selected human labeling (ablation) | TMA incremental updates · VIPeR | rank-1 accuracy | Hurt |
Closest papers by meaning
- Distilled Person Re-Identification: Towards a More Scalable System (CVPR 2019 · similarity 0.89)
- Interpretable and Generalizable Person Re-Identification with Query-Adaptive Convolution and Temporal Lifting (ECCV 2020 · similarity 0.89)
- Efficient Online Local Metric Adaptation via Negative Samples for Person Re-Identification (ICCV 2017 · similarity 0.89)
- Person Re-Identification by Camera Correlation Aware Feature Augmentation (TPAMI 2017 · similarity 0.89)
- Unsupervised Adaptive Re-Identification in Open World Dynamic Camera Networks (CVPR 2017 · similarity 0.88)
- Learning Instance-Level Spatial-Temporal Patterns for Person Re-Identification (ICCV 2021 · similarity 0.88)
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.
