Research map / Detection and segmentation / CNN features
Anchor-Free 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 |
|---|---|---|---|
| Use a single P3 feature level rather than multiple levels (ablation) | AlignPS · CUHK-SYSU | mAP: 90.0 -> 93.1 | Helped |
| Add lateral deformable convolution (ablation) | AlignPS AFA · CUHK-SYSU | mAP: 83.4 -> 90.6 | Helped |
| Add output deformable convolution (ablation) | AlignPS AFA · CUHK-SYSU | mAP: 83.4 -> 91.4 | Helped |
| Add feature concatenation (ablation) | AlignPS AFA · CUHK-SYSU | mAP: 83.4 -> 84.0 | Helped |
| Combine lateral deformable convolution, output deformable convolution, and feature concatenation (ablation) | AlignPS AFA · CUHK-SYSU | mAP: 83.4 -> 93.1 | Helped |
| Make re-id primary instead of sharing its features with the detection head (ablation) | AlignPS · CUHK-SYSU | mAP: 80.3 -> 93.1 | Helped |
| Make re-id primary instead of using independent detection and re-id heads (ablation) | AlignPS · CUHK-SYSU | mAP: 80.7 -> 93.1 | Helped |
| Add triplet loss to OIM without lookup-table features (ablation) | AlignPS re-id loss · CUHK-SYSU | mAP: 92.4 -> 92.8 | Helped |
| Include lookup-table features in triplet-aided OIM (ablation) | AlignPS re-id loss · CUHK-SYSU | mAP: 92.4 -> 93.1; +0.7 | Helped |
| Add deformable convolutions in res3, res4 and res5 of the backbone (ablation) | AlignPS · CUHK-SYSU | mAP: 93.1 -> 94.0 | Helped |
| Use multiple feature levels instead of P3 only (ablation) | AlignPS · CUHK-SYSU | Recall: 90.3 -> 90.9 | Mixed |
| Use AlignPS rather than NAE+ | Person search · CUHK-SYSU | mAP: 92.1 -> 93.1 | Helped |
Closest papers by meaning
- Exploring Visual Context for Weakly Supervised Person Search (AAAI 2022 · similarity 0.87)
- FSLoRA: Harmonizing Detection and Re-Identification via Freq-Spatial Low-Rank Adapter for One-Stage Person Search (CVPR 2026 · similarity 0.86)
- Diverse Knowledge Distillation for End-to-End Person Search (AAAI 2021 · similarity 0.86)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · similarity 0.84)
- Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · similarity 0.83)
- Automatic Synthetic Data and Fine-grained Adaptive Feature Alignment for Composed Person Retrieval (NeurIPS 2025 · similarity 0.83)
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.