Research map / Label-efficient and robust learning / Recognition and long-tail
Spectral Feature Transformation for Person Re-Identification
Chuanchen Luo, Yuntao Chen, Naiyan Wang, Zhaoxiang Zhang
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 spectral feature transformation (SFT) to the baseline (ablation) | AM-Softmax ResNet-50 baseline · CUHK03 (labeled) | mAP / Rank-1: 40.6 -> 60.2 mAP; 44.9 -> 66.3 Rank-1 | Improved |
| add SFT to the baseline (ablation) | AM-Softmax ResNet-50 baseline · Market-1501 | mAP / Rank-1: 77.3 -> 79.6 mAP; 91.2 -> 91.6 Rank-1 | Improved |
| add SFT to the baseline (ablation) | AM-Softmax ResNet-50 baseline · MSMT17 | mAP / Rank-1: 37.3 -> 44.7 mAP; 66.7 -> 71.9 Rank-1 | Improved |
| add deep supervision with a shared classifier on the original feature (ablation) | SFT · Market-1501 | mAP / Rank-1: 79.6 -> 82.7 mAP; 91.6 -> 93.4 Rank-1 | Improved |
| add deep supervision with a shared classifier on the original feature (ablation) | SFT · MSMT17 | mAP / Rank-1: 44.7 -> 47.6 mAP; 71.9 -> 73.6 Rank-1 | Improved |
| share the classifier between original and transformed features instead of using an unshared one (ablation) | SFT + deep supervision · MSMT17 | mAP / Rank-1: 35.3 -> 47.6 mAP; 63.7 -> 73.6 Rank-1 | Improved |
| share the classifier between original and transformed features instead of using an unshared one (ablation) | SFT + deep supervision · Market-1501 | mAP / Rank-1: 79.0 -> 82.7 mAP; 91.8 -> 93.4 Rank-1 | Improved |
| add proposed top-n post-processing (ablation) | SFT + deep supervision · Market-1501 | mAP / Rank-1: 82.7 -> 87.5 mAP; 93.4 -> 94.1 Rank-1 | Improved |
| add proposed top-n post-processing (ablation) | SFT + deep supervision · CUHK03 (labeled) | mAP / Rank-1: 62.4 -> 71.7 mAP; 68.2 -> 74.3 Rank-1 | Improved |
| use k-reciprocal re-ranking instead of proposed post-processing | SFT + deep supervision · Market-1501 | mAP / Rank-1: 87.5 -> 90.6 mAP; 94.1 -> 93.5 Rank-1 | Mixed |
| replace Ncut loss with SFT | same ResNet-50 backbone · CUHK03 (labeled) | mAP / Rank-1: 40.3 -> 62.4 mAP; 45.5 -> 68.2 Rank-1 | Improved |
| replace Ncut loss with SFT | same ResNet-50 backbone · Market-1501 | mAP / Rank-1: 79.5 -> 82.7 mAP; 92.0 -> 93.4 Rank-1 | Improved |
Closest papers by meaning
- Embedding Deep Metric for Person Re-identification: A Study Against Large Variations (ECCV 2016 · similarity 0.88)
- Deep Neural Networks with Inexact Matching for Person Re-Identification (NeurIPS 2016 · similarity 0.88)
- Resource Aware Person Re-Identification Across Multiple Resolutions (CVPR 2018 · similarity 0.87)
- A Bottom-Up Clustering Approach to Unsupervised Person Re-Identification (AAAI 2019 · similarity 0.87)
- Asymmetric Co-Teaching for Unsupervised Cross-Domain Person Re-Identification (AAAI 2020 · similarity 0.87)
- Deep Learning for Person Re-Identification: A Survey and Outlook (TPAMI 2021 · similarity 0.87)
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.
