Research map / Video / Video segmentation and tracking
Dense Interaction Learning for Video-Based Person Re-Identification
Tianyu He, Xin Jin, Xu Shen, Jianqiang Huang, Zhibo Chen, Xian-Sheng Hua
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 Dense Attention in the decoder (vs plain CNN baseline) (ablation) | ResNet-50 baseline · MARS | mAP: 82.1 -> 87.0 | Improved |
| add self-attention decoder (CNN-TransEnc) (ablation) | ResNet-50 baseline · MARS | mAP: +3.6% mAP | Improved |
| use Dense Attention instead of dense summation or concatenation (ablation) | DenseIL · MARS | mAP: 86.2 -> 87.0 | Improved |
| use Dense Attention instead of encoder-decoder attention on top-level features (ablation) | CNN-TransDec · MARS | mAP: 85.8 -> 87.0 | Improved |
| add STEP-Emb positional embedding (ablation) | DenseIL · MARS | mAP: 86.3 -> 87.0 | Improved |
| add only spatial or temporal positional embedding (ablation) | DenseIL · MARS | mAP: 86.3 -> 86.6 | Improved |
| increase number of decoder blocks from 0 to 4 (ablation) | DenseIL · MARS | mAP: 82.1 -> 87.0 | Improved |
| widen hidden dimension from 256 to 2048 (ablation) | DenseIL · MARS | mAP: 86.6 -> 87.0 | Improved |
| add the lowest-level CNN feature Z1 to Dense Attention (ablation) | DenseIL · MARS | mAP: 87.0 -> 86.9 | Hurt |
| add intermediate CNN feature Z3 to Z4 in Dense Attention (ablation) | DenseIL · MARS | mAP: 85.8 -> 86.7 | Improved |
| increase partitions from 4 to 8 (ablation) | DenseIL · MARS | mAP: 87.0 -> 86.6 | Hurt |
| increase partitions from 1 to 4 (ablation) | DenseIL · MARS | mAP: 86.1 -> 87.0 | Improved |
Closest papers by meaning
- Top-Push Video-Based Person Re-Identification (CVPR 2016 · similarity 0.89)
- Deep Learning for Person Re-Identification: A Survey and Outlook (TPAMI 2021 · similarity 0.88)
- Video-Based Person Re-Identification With Spatial and Temporal Memory Networks (ICCV 2021 · similarity 0.88)
- Event-Guided Person Re-Identification via Sparse-Dense Complementary Learning (CVPR 2023 · similarity 0.88)
- Relation-Guided Spatial Attention and Temporal Refinement for Video-Based Person Re-Identification (AAAI 2020 · similarity 0.88)
- Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-Identification (CVPR 2018 · 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.
