Research map / Video / Video-language understanding
Online Multi-modal Person Search in Videos
Jiangyue Xia, Anyi Rao, Qingqiu Huang, Linning Xu, Jiangtao Wen, Dahua Lin
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 reinforcement-learned controller instead of manual threshold rule (ablation) | OMS dynamic memory bank · CSM test (per movie) | mAP (%): 63.83 -> 64.39 | Improved |
| add body and audio multi-modal features (ablation) | OMS-R · CSM test (per movie) | mAP (%): 64.39 -> 66.42 | Improved |
| add uncertain instance cache (ablation) | OMS-RM · CSM test (per movie) | mAP (%): 66.42 -> 69.08 | Improved |
| add dynamic memory bank updating to face matching (ablation) | face matching · CSM test (per movie) | mAP (%): 61.24 -> 63.83 | Improved |
| add body feature to face (ablation) | OMS (DMB w/ RLC+UIC) · CSM test (per movie) | mAP (%): 64.91 -> 67.93 | Improved |
| add audio feature to face (ablation) | OMS (DMB w/ RLC+UIC) · CSM test (per movie) | mAP (%): 64.91 -> 65.39 | Improved |
| add all three modalities versus face only (ablation) | OMS (DMB w/ RLC+UIC) · CSM test (per movie) | mAP (%): 64.91 -> 69.08 | Improved |
| raise cache weight from 0 to 0.08 (smaller cache) (ablation) | OMS-RMQ · CSM test (per movie) | mAP (%): 66.84 -> 69.08 | Improved |
| raise cache weight from 0.08 to 0.20 (cache too small) (ablation) | OMS-RMQ · CSM test (per movie) | mAP (%): 69.08 -> 63.69 | Hurt |
| online OMS-RMQ versus offline PPCC | OMS-RMQ vs PPCC · CSM test (per movie) | mAP (%): 67.99 -> 69.08 | Improved |
| add cache over time (R@1 along the movie) (ablation) | OMS-RMQ · CSM test (per movie) | R@1: 59% -> 67% | Improved |
Closest papers by meaning
- Temporal Complementarity-Guided Reinforcement Learning for Image-to-Video Person Re-Identification (CVPR 2022 · similarity 0.87)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · similarity 0.85)
- Person Search in Videos with One Portrait Through Visual and Temporal Links (ECCV 2018 · similarity 0.85)
- Video-Based Person Re-Identification With Spatial and Temporal Memory Networks (ICCV 2021 · similarity 0.85)
- Deep Reinforcement Active Learning for Human-in-the-Loop Person Re-Identification (ICCV 2019 · similarity 0.84)
- TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · 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.
