Research map / Video / Action recognition
ActionSwitch: Class-agnostic Detection of Simultaneous Actions in Streaming Videos
Hyolim Kang, Jeongseok Hyun, Joungbin An, Youngjae Yu, Seon Joo Kim
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 the conservativeness loss (alpha=0.025) to a 2-switch ActionSwitch (ablation) | ActionSwitch · THUMOS14 On-TAL | F1: 45.87 -> 53.20 | Improved |
| add the conservativeness loss (alpha=0.010) to a 2-switch ActionSwitch (ablation) | ActionSwitch · FineAction On-TAL | F1: 17.57 -> 19.44 | Improved |
| increase conservativeness weight alpha beyond the best value (ablation) | ActionSwitch 2 switches · THUMOS14 On-TAL | F1: 53.20 -> 50.20 | Hurt |
| increase alpha from 0.010 to 0.025 on FineAction (ablation) | ActionSwitch 2 switches · FineAction On-TAL | F1: 19.44 -> 18.71 | Hurt |
| use 2 switches instead of 1 switch (ablation) | ActionSwitch · THUMOS14 On-TAL | recall: 59.49 -> 63.96 | Improved |
| use 2 switches instead of 1 switch (ablation) | ActionSwitch · FineAction On-TAL | recall: 14.71 -> 22.36 | Improved |
| use 3 switches instead of 2 on densely overlapping MultiTHUMOS (ablation) | ActionSwitch · MultiTHUMOS On-TAL | F1: 30.82 -> 32.76 | Improved |
| use 4 switches instead of 3 on MultiTHUMOS (ablation) | ActionSwitch · MultiTHUMOS On-TAL | F1: 32.76 -> 29.25 | Hurt |
| add conservativeness loss to 3-switch ActionSwitch (ablation) | ActionSwitch · MultiTHUMOS On-TAL | F1: 26.63 -> 32.76 | Improved |
| use ActionSwitch instead of OAT on THUMOS14 | On-TAL · THUMOS14 On-TAL | F1: 62.9 -> 53.2 | Hurt |
| use ActionSwitch instead of OAT on Epic-Kitchens 100 | On-TAL · Epic-Kitchens 100 (noun) | F1: 27.583 -> 32.444 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CAG-QIL — CAG-QIL: Context-Aware Actionness Grouping via Q Imitation Learning for Online Temporal Action Localization (ICCV 2021) · better on 4 of 4 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
- Watch Only Once: An End-to-End Video Action Detection Framework (ICCV 2021 · similarity 0.91)
- Action Sensitivity Learning for Temporal Action Localization (ICCV 2023 · similarity 0.90)
- OnPoint: Offline-to-Online Multi-Level Distillation for Point-Supervised Online Temporal Action Localization (ECCV 2026 · similarity 0.90)
- Learning To Refactor Action and Co-Occurrence Features for Temporal Action Localization (CVPR 2022 · similarity 0.89)
- Learning Temporal Co-Attention Models for Unsupervised Video Action Localization (CVPR 2020 · similarity 0.89)
- ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization (AAAI 2021 · similarity 0.89)
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.
