Research map / Video / Action recognition
OpenTAL: Towards Open Set Temporal Action Localization
Wentao Bao, Qi Yu, Yu Kong
Check 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 OpenTAL instead of vanilla EDL uncertainty scoring | AFSD · THUMOS14 unknown | OSDR: 36.26 -> 42.91 | Improved |
| Use OpenTAL instead of vanilla EDL uncertainty scoring | AFSD · ActivityNet1.3 unknown | OSDR: 38.56 -> 50.49 | Improved |
| Add MIB importance re-weighting (ablation) | OpenTAL · THUMOS14 unknown | OSDR: 12.10 -> 42.91 | Improved |
| Add actionness prediction instead of using (K+1)-way classification (ablation) | OpenTAL · THUMOS14 unknown | AUROC: 58.12 -> 78.33 | Improved |
| Add actionness prediction instead of using (K+1)-way classification (ablation) | OpenTAL · THUMOS14 unknown | FAR@95: 82.85 -> 70.96 | Improved |
| Add IoU-aware uncertainty calibration (ablation) | OpenTAL · THUMOS14 unknown | AUPR: 37.86 -> 58.62 | Improved |
| Replace soft MIB weight updates with hard updates (epsilon=0) (ablation) | OpenTAL · THUMOS14 unknown | FAR@95: 70.96 -> 66.34 | Improved |
| Replace soft MIB weight updates with hard updates (epsilon=0) (ablation) | OpenTAL · THUMOS14 unknown | OSDR: 42.91 -> 38.90 | Hurt |
| Use focal loss instead of MIB re-weighting (ablation) | OpenTAL · THUMOS14 unknown | OSDR: 42.91 -> 2.04 | Hurt |
| Use GHM instead of MIB re-weighting (ablation) | OpenTAL · THUMOS14 unknown | OSDR: 42.91 -> 1.41 | Hurt |
| Use IB instead of MIB re-weighting (ablation) | OpenTAL · THUMOS14 unknown | OSDR: 42.91 -> 2.18 | Hurt |
| Use the two-level actionness-then-uncertainty decision instead of multiplying actionness and uncertainty (ablation) | OpenTAL inference · THUMOS14 unknown | OSDR: 42.44 -> 42.91 | Improved |
Closest papers by meaning
- Cascade Evidential Learning for Open-World Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.90)
- Vectorized Evidential Learning for Weakly-Supervised Temporal Action Localization (TPAMI 2023 · similarity 0.89)
- Dual-Evidential Learning for Weakly-Supervised Temporal Action Localization (ECCV 2022 · similarity 0.88)
- Test-Time Zero-Shot Temporal Action Localization (CVPR 2024 · similarity 0.87)
- Energy vs. Noise: Towards Robust Temporal Action Localization in Open-World (AAAI 2025 · similarity 0.87)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.86)
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.
