Research map / Label-efficient and robust learning / Recognition and long-tail
Batch DropBlock Network for Person Re-Identification and Beyond
Zuozhuo Dai, Mingqiang Chen, Xiaodong Gu, Siyu Zhu, Ping Tan
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 feature dropping branch (Batch DropBlock) (ablation) | Baseline + Triplet · CUHK03-Detect | Rank-1: 63.6 -> 72.8 | Improved |
| add feature dropping branch (Batch DropBlock) (ablation) | Baseline + Triplet · Market-1501 | Rank-1: 93.1 -> 94.2 | Improved |
| add global branch to feature dropping branch (ablation) | BDB · Market-1501 | Rank-1: 93.6 -> 94.2 | Improved |
| add triplet loss to Baseline + Dropping (ablation) | BDB · CUHK03-Detect | Rank-1: 60.9 -> 72.8 | Improved |
| use Batch DropBlock instead of DropBlock (ablation) | BDB · CUHK03-Detect | Rank-1: 70.6 -> 72.8 | Improved |
| use Batch DropBlock instead of Batch Dropout (ablation) | BDB · CUHK03-Detect | Rank-1: 65.8 -> 72.8 | Improved |
| add Cutout to feature dropping branch without global branch (ablation) | BDB · Market-1501 | Rank-1: 93.6 -> 88.0 | Hurt |
| add Cutout augmentation to BDB (ablation) | BDB · Market-1501 | Rank-1: 94.2 -> 95.3 | Improved |
| add Random Erasing to BDB (ablation) | BDB · CUHK03-Detect | Rank-1: 72.8 -> 75.9 | Improved |
| add Cutout to the baseline (ablation) | Baseline + Triplet · CUHK03-Detect | Rank-1: 63.6 -> 67.7 | Improved |
| use global max pooling instead of average pooling in dropping branch (ablation) | BDB · CUHK03-Detect | Rank-1 | Improved |
| set erased height ratio to 0.3 (ablation) | BDB · CUHK03-Detect | Rank-1 | Improved |
Closest papers by meaning
- Second-Order Non-Local Attention Networks for Person Re-Identification (ICCV 2019 · similarity 0.88)
- Learning Deep Feature Representations With Domain Guided Dropout for Person Re-Identification (CVPR 2016 · similarity 0.87)
- Learning Generalisable Omni-Scale Representations for Person Re-Identification (TPAMI 2021 · similarity 0.86)
- Omni-Scale Feature Learning for Person Re-Identification (ICCV 2019 · similarity 0.86)
- Deep Neural Networks with Inexact Matching for Person Re-Identification (NeurIPS 2016 · similarity 0.86)
- Keypoint Message Passing for Video-Based Person Re-identification (AAAI 2022 · 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.
