Research map / Video / Action recognition
Temporal Action Localization With Pyramid of Score Distribution Features
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 |
|---|---|---|---|
| represent action evidence with multi-resolution PSDF and classify its score distribution using Temporal-SVM | temporal action localization · THUMOS'14 | mAP: 36.6 -> 51.4 at overlap 0.1; 14.4 -> 18.8 at overlap 0.5 | Helped |
| use multi-resolution PSDF with Temporal-SVM | temporal action localization · THUMOS'15 | mAP | Helped |
| replace max-score classification with Temporal-SVM over the PSDF distribution (ablation) | PSDF action detection · THUMOS'14 | mAP: 45.6 -> 51.4 at overlap 0.1; 16.7 -> 18.8 at overlap 0.5 | Helped |
| add Elman-Net temporal modeling to PSDF (ablation) | PSDF action detection · THUMOS'14 | mAP: 51.4 -> 62.1 at overlap 0.1; 18.8 -> 16.3 at overlap 0.5 | Mixed |
| add LSTM temporal modeling to PSDF (ablation) | PSDF action detection · THUMOS'14 | mAP: 51.4 -> 58.8 at overlap 0.1; 18.8 -> 16.0 at overlap 0.5 | Mixed |
| use sequence length 20 rather than 40 for Elman-Net (ablation) | PSDF + Elman-Net · THUMOS'14 | mAP: 60.8 -> 62.1 at overlap 0.1; 16.5 -> 16.3 at overlap 0.5 | Mixed |
| add PSDF with Temporal-SVM | MPII Cooking detection · MPII Cooking, first subject | precision: 17.7 -> 27.3 | Helped |
| add recurrent temporal modeling to PSDF (ablation) | MPII Cooking detection · MPII Cooking, first subject | recall: 42.5 -> 56.3 (Elman-Net); 42.5 -> 59.7 (LSTM) | Helped |
| add smoothing regularization on network input weights (ablation) | PSDF detection networks · detection experiments | performance: about 1% | Helped |
| balance classes using weights inversely proportional to action instances (ablation) | Temporal-SVM and RNN · detection task | classifier performance | Helped |
Closest papers by meaning
- Three Birds with One Stone: Multi-Task Temporal Action Detection via Recycling Temporal Annotations (CVPR 2021 · similarity 0.89)
- Temporal Action Localization in Untrimmed Videos via Multi-Stage CNNs (CVPR 2016 · similarity 0.89)
- Learning To Refactor Action and Co-Occurrence Features for Temporal Action Localization (CVPR 2022 · similarity 0.89)
- Rethinking the Faster R-CNN Architecture for Temporal Action Localization (CVPR 2018 · similarity 0.88)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.88)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.87)
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.