Research map / Detection and segmentation / CNN features
Cascade Transformers for End-to-End Person Search
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the three-stage cascade instead of using one stage (ablation) | COAT with conv5 blocks · PRW | mAP: 43.5 -> 49.5 | Helped |
| add the occluded attention transformers instead of conv5 blocks (ablation) | three-stage COAT · PRW | mAP: 49.5 -> 53.3 | Helped |
| remove ReID loss at stage 1 (ablation) | COAT · PRW | mAP: 50.3 -> 53.3 | Helped |
| add progressively increasing IoU thresholds 0.5, 0.6, 0.7 instead of equal 0.5 thresholds (ablation) | COAT · PRW | mAP: 52.5 -> 53.3 | Helped |
| add progressively increasing IoU thresholds 0.5, 0.6, 0.7 instead of equal 0.6 thresholds (ablation) | COAT · PRW | top-1: 86.2 -> 87.4 | Helped |
| replace occluded attention with vanilla attention (ablation) | COAT · PRW | top-1: 86.4 -> 87.4 | Hurt |
| replace occluded attention with CrossViT (ablation) | COAT · PRW | mAP: 49.9 vs. 53.3 | Hurt |
| replace occluded attention with Jigsaw (ablation) | COAT · PRW | mAP: 51.9 vs. 53.3 | Hurt |
| replace token-level occluded attention with Batch DropBlock feature augmentation (ablation) | COAT · PRW | mAP: 52.7 vs. 53.3 | Hurt |
| replace token-level occluded attention with Cutout feature augmentation (ablation) | COAT · PRW | mAP: 53.2 vs. 53.3 | Hurt |
| replace token-level occluded attention with Mixup feature augmentation (ablation) | COAT · PRW | mAP: 52.8 vs. 53.3 | Hurt |
| use multiple convolutional tokenization scales, 1x1 and 3x3, instead of only 1x1 (ablation) | COAT with occluded attention · PRW | mAP: 52.2 -> 53.3 | Helped |
Closest papers by meaning
- Diverse Knowledge Distillation for End-to-End Person Search (AAAI 2021 · similarity 0.87)
- Norm-Aware Embedding for Efficient Person Search (CVPR 2020 · similarity 0.86)
- Pose Recognition With Cascade Transformers (CVPR 2021 · similarity 0.86)
- TransMatcher: Deep Image Matching Through Transformers for Generalizable Person Re-identification (NeurIPS 2021 · similarity 0.85)
- End-to-End Thorough Body Perception for Person Search (AAAI 2020 · similarity 0.85)
- Attention Disturbance and Dual-Path Constraint Network for Occluded Person Re-identification (AAAI 2024 · similarity 0.84)
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.