Research map / Video / Action recognition
Online Action Detection
Roeland De Geest, Efstratios Gavves, Amir Ghodrati, Zhenyang Li, Cees Snoek, Tinne Tuytelaars
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 |
|---|---|---|---|
| use Fisher vectors on improved trajectories instead of per-frame CNN | online action detector · TVSeries online | mAP (%): 1.9 -> 5.2 | Improved |
| add LSTM temporal modelling on top of per-frame CNN features | CNN baseline · TVSeries online | mAP (%): 1.9 -> 2.7 | Improved |
| use FV instead of LSTM | online action detector · TVSeries online | mAP (%): 2.7 -> 5.2 | Improved |
| use FV instead of CNN | offline action detector · TVSeries offline (overlap 0.2) | mAP (%): 1.1 -> 4.9 | Improved |
| view only the first 10% instead of the last 10% of the action | FV · TVSeries online | mean cAP (%): 76.8 -> 67.0 | Hurt |
| view only the first 10% instead of the last 10% of the action | CNN · TVSeries online | mean cAP (%): 61.5 -> 61.0 | No clear effect |
| view only the first 10% instead of the last 10% of the action | LSTM · TVSeries online | mean cAP (%): 64.3 -> 63.3 | No clear effect |
| small or background action instead of regular (ablation) | FV · TVSeries metadata | mean cAP difference: -5.8 | Hurt |
| small or background action instead of regular (ablation) | CNN · TVSeries metadata | mean cAP difference: 3.0 | Improved |
| multiple persons in scene (ablation) | CNN · TVSeries metadata | mean cAP difference: -3.2 | Hurt |
| multiple persons in scene (ablation) | FV · TVSeries metadata | mean cAP difference: 2.8 | Improved |
| shotcut during the action (ablation) | FV and LSTM versus CNN · TVSeries metadata | mean cAP difference: FV -0.8, CNN 1.7, LSTM 5.4 | Mixed |
Closest papers by meaning
- MiniROAD: Minimal RNN Framework for Online Action Detection (ICCV 2023 · similarity 0.88)
- ActionSwitch: Class-agnostic Detection of Simultaneous Actions in Streaming Videos (ECCV 2024 · similarity 0.88)
- Learning to Discriminate Information for Online Action Detection (CVPR 2020 · similarity 0.87)
- Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection? (NeurIPS 2024 · similarity 0.86)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.85)
- Benchmarking the Robustness of Temporal Action Detection Models Against Temporal Corruptions (CVPR 2024 · similarity 0.85)
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.
