Research map / Video / Action recognition
Temporal Action Detection With Structured Segment Networks
Yue Zhao, Yuanjun Xiong, Limin Wang, Zhirong Wu, Xiaoou Tang, 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 structured temporal pyramid pooling with stage structure (A=1 instead of A=0) (ablation) | SSN avg pool · ActivityNet v1.2 | average mAP (%): 4.34 -> 24.6 | Improved |
| use average pooling instead of max pooling (ablation) | SSN (1,2)-1 · ActivityNet v1.2 | average mAP (%): 18.4 -> 24.6 | Improved |
| increase pyramid depth from (1) to (1,2) (ablation) | SSN avg pool A=1 · ActivityNet v1.2 | average mAP (%): 24.3 -> 24.6 | Improved |
| add separate activity and completeness classifiers instead of single classifier (ablation) | SSN · ActivityNet v1.2 | mAP (%): 17.9% -> 23.7% | Improved |
| add location regression with multi-task learning (end-to-end) (ablation) | SSN · ActivityNet v1.2 | mAP (%): 24.8 -> 25.9 | Improved |
| add location regression (stage-wise) (ablation) | SSN · ActivityNet v1.2 | mAP (%): 24.6 -> 24.9 | Improved |
| train end-to-end instead of stage-wise (ablation) | SSN · ActivityNet v1.2 | mAP (%): 24.9 -> 25.9 | Improved |
| use TAG proposals instead of sliding windows (ablation) | SSN · ActivityNet v1.2 | mAP (%): 18.1 -> 24.9 | Improved |
| use TAG proposals instead of sliding windows (ablation) | proposal generation · THUMOS14 | average recall (%): 21.2 -> 48.9 | Improved |
| use TAG proposals instead of sliding windows (ablation) | proposal generation · ActivityNet v1.2 | average recall (%): 34.8 -> 71.7 | Improved |
| apply actionness classifier trained on ActivityNet v1.2 directly to THUMOS14 | TAG · THUMOS14 | average recall (%): 48.9 -> 39.6 | Hurt |
| add SSN structured detection compared with prior best | SSN · THUMOS14 | mAP@0.5 (%): 19.0 -> 29.1 | Improved |
Closest papers by meaning
- Temporal Segment Networks for Action Recognition in Videos (TPAMI 2018 · similarity 0.91)
- MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation (CVPR 2019 · similarity 0.90)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.90)
- Action Sets: Weakly Supervised Action Segmentation Without Ordering Constraints (CVPR 2018 · similarity 0.90)
- Deep Learning-Based Action Detection in Untrimmed Videos: A Survey (TPAMI 2022 · similarity 0.89)
- Completeness Modeling and Context Separation for Weakly Supervised Temporal Action Localization (CVPR 2019 · 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.
