Research map / Video / Video segmentation and tracking
Reinforced Temporal Attention and Split-Rate Transfer for Depth-Based Person Re-Identification
Nikolaos Karianakis, Zicheng Liu, Yinpeng Chen, Stefano Soatto
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 Reinforced Temporal Attention (ablation) | CNN-LSTM+Avg Pooling · DPI-T (multi-shot) | top-1 accuracy (%): 75.5 -> 76.3 | Improved |
| add Reinforced Temporal Attention (ablation) | CNN-LSTM+Avg Pooling · BIWI (multi-shot) | top-1 accuracy (%): 45.7 -> 50.0 | Improved |
| add Reinforced Temporal Attention (ablation) | CNN-LSTM+Avg Pooling · IIT PAVIS (multi-shot) | top-1 accuracy (%): 50.1 -> 52.4 | Improved |
| use RTA instead of soft attention (ablation) | CNN-LSTM · DPI-T (multi-shot) | top-1 accuracy (%): 75.9 -> 76.3 | Improved |
| use RTA instead of soft attention (ablation) | CNN-LSTM · BIWI (multi-shot) | top-1 accuracy (%): 46.4 -> 50.0 | Improved |
| use RTA instead of soft attention (ablation) | CNN-LSTM · IIT PAVIS (multi-shot) | top-1 accuracy (%): 50.6 -> 52.4 | Improved |
| add LSTM over frame features with average pooling (multi-shot) | CNN · DPI-T | top-1 accuracy (%): 66.8 -> 75.5 | Improved |
| add LSTM over frame features with average pooling (multi-shot) | CNN · BIWI | top-1 accuracy (%): 25.4 -> 45.7 | Improved |
| add LSTM over frame features with average pooling (multi-shot) | CNN · IIT PAVIS | top-1 accuracy (%): 43.0 -> 50.1 | Improved |
| transfer and freeze bottom 3 RGB layers, rapidly fine-tune top layers (split-rate) vs Yosinski transfer (ablation) | CNN embedding · DPI-T | top-1 accuracy (%) | Improved |
| use 3D RAM with spatial attention instead of our CNN (single-shot) | Our method (CNN) · BIWI | top-1 accuracy (%): 30.1 vs 25.4 | Hurt |
Closest papers by meaning
- Recurrent Attention Models for Depth-Based Person Identification (CVPR 2016 · similarity 0.89)
- Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-Identification (CVPR 2018 · similarity 0.88)
- Temporal-Enhanced Convolutional Network for Person Re-Identification (AAAI 2018 · similarity 0.88)
- Recurrent Convolutional Network for Video-Based Person Re-Identification (CVPR 2016 · similarity 0.88)
- Leader-Based Multi-Scale Attention Deep Architecture for Person Re-Identification (TPAMI 2019 · similarity 0.88)
- Co-Segmentation Inspired Attention Networks for Video-Based Person Re-Identification (ICCV 2019 · 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.
