Research map / Video / Video segmentation and tracking
READ: Reciprocal Attention Discriminator for Image-to-Video Re-Identification
Minho Shim, Hsuan-I Ho, Jinhyung Kim, Dongyoon Wee
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 reciprocal attention (READ) over the discriminator baseline (ablation) | baseline (discriminator) · DukeMTMC-VideoReID | top-1 accuracy: 75.2 -> 84.1 | Improved |
| add reciprocal attention (READ) over the discriminator baseline (ablation) | baseline (discriminator) · MARS | top-1 accuracy: 65.0 -> 80.3 | Improved |
| replace metric learning with a discriminator (ablation) | metric baseline · DukeMTMC-VideoReID | top-1 accuracy: 67.5 -> 75.2 | Improved |
| replace metric learning with a discriminator (ablation) | metric baseline · MARS | top-1 accuracy: 67.1 -> 65.0 | Hurt |
| add reciprocal attention triplet loss (RATL) (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 84.1 -> 86.3 | Improved |
| add reciprocal attention triplet loss (RATL) (ablation) | READ · MARS | top-1 accuracy: 80.3 -> 81.5 | Improved |
| add random horizontal flip (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 85.8 -> 86.3 | No clear effect |
| reduce #avgID from 32 to 16 (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 86.3 -> 86.0 | No clear effect |
| reduce #avgID from 32 to 16 (ablation) | READ · MARS | top-1 accuracy: 81.5 -> 76.6 | Hurt |
| add query sample normalization (multiple query frames averaged) (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 80.3 -> 86.3 | Improved |
| increase gallery sample length T from 2 to 4 (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 83.2 -> 86.3 | Improved |
| increase gallery sample length T from 4 to 6 (ablation) | READ · DukeMTMC-VideoReID | top-1 accuracy: 86.3 -> 85.2 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports READ as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Online Reasoning Video Object Segmentation (ECCV 2026 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Watching You: Global-Guided Reciprocal Learning for Video-Based Person Re-Identification (CVPR 2021 · similarity 0.90)
- Scalable Person Re-Identification by Harmonious Attention (IJCV 2019 · similarity 0.89)
- Person Re-Identification by Camera Correlation Aware Feature Augmentation (TPAMI 2017 · similarity 0.89)
- A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets (TPAMI 2018 · similarity 0.89)
- Co-Segmentation Inspired Attention Networks for Video-Based Person Re-Identification (ICCV 2019 · similarity 0.88)
- Temporal Correlation Vision Transformer for Video Person Re-Identification (AAAI 2024 · 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.
