Research map / Label-efficient and robust learning / Recognition and long-tail
Person Re-identification via Recurrent Feature Aggregation
Yichao Yan, Bingbing Ni, Zhichao Song, Chao Ma, Yan Yan, Xiaokang Yang
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 LSTM recurrent aggregation (last node instead of first node features) (ablation) | RFA-Net · iLIDS-VID and PRID 2011 | rank-1 matching rate: about 10% gain | Improved |
| use RFA-Net+RSVM instead of Color&LBP+DVR | Color&LBP features · iLIDS-VID | rank-1 matching rate (%): 34.5 -> 49.3 | Improved |
| use RFA-Net+RSVM instead of Color&LBP+DVR | Color&LBP features · PRID 2011 | rank-1 matching rate (%): 37.6 -> 58.2 | Improved |
| use RFA-Net+RSVM instead of averaged-pool Color&LBP+RSVM | Color&LBP features · iLIDS-VID | rank-1 matching rate (%): 23.2 -> 49.3 | Improved |
| add RankSVM metric learning over cosine distance (ablation) | RFA-Net · iLIDS-VID | rank-1 matching rate (%): 44.5 -> 49.3 | Improved |
| add RankSVM metric learning over cosine distance (ablation) | RFA-Net · PRID 2011 | rank-1 matching rate (%): 54.9 -> 58.2 | Improved |
| add RankSVM over cosine distance (ablation) | RFA-Net · iLIDS-VID rank-20 | rank-20 matching rate (%): 90.1 -> 90.0 | No clear effect |
| add 50% noise frames to test sequences | RFA-Net+RSVM · iLIDS-VID | rank-1 matching rate (%): 49.3 -> 29.8 | Hurt |
| add 50% noise frames to test sequences | RFA-Net+RSVM · PRID 2011 | rank-1 matching rate (%): 58.2 -> 44.7 | Hurt |
| average features over 10 test subsequences instead of 1 (ablation) | RFA-Net+RSVM · iLIDS-VID | rank-1 matching rate (%): 33.1 -> 49.3 | Improved |
| average features over 10 test subsequences instead of 1 (ablation) | RFA-Net+RSVM · PRID 2011 | rank-1 matching rate (%): 45.6 -> 58.2 | Improved |
| average 15 test subsequences instead of 10 (ablation) | RFA-Net+RSVM · iLIDS-VID | rank-1 matching rate (%): 49.3 -> 47.8 | No clear effect |
Closest papers by meaning
- Interaction-And-Aggregation Network for Person Re-Identification (CVPR 2019 · similarity 0.89)
- A Siamese Long Short-Term Memory Architecture for Human Re-identification (ECCV 2016 · similarity 0.88)
- Recurrent Convolutional Network for Video-Based Person Re-Identification (CVPR 2016 · similarity 0.88)
- Ordered or Orderless: A Revisit for Video Based Person Re-Identification (TPAMI 2020 · similarity 0.88)
- Towards Discriminative Representation Learning for Unsupervised Person Re-Identification (ICCV 2021 · similarity 0.88)
- See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 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.
