Research map / Video / Action recognition
SSNet: Scale Selection Network for Online 3D Action Prediction
Jun Liu, Amir Shahroudy, Gang Wang, Ling-Yu Duan, Alex C. Kot
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 |
|---|---|---|---|
| Select the temporal window scale dynamically using the previous step's regressed action-start distance | SSNet · OAD | accuracy: 65.6% vs 62.0%; 79.2% vs 77.2%; 81.6% vs 79.7% | Improved |
| Select the temporal window scale dynamically using the previous step's regressed action-start distance | SSNet · PKU-MMD | accuracy: 30.0% vs 25.6%; 68.5% vs 66.4%; 78.6% vs 75.5% | Improved |
| Use regressed scales instead of ground-truth scales for selection | SSNet · OAD and PKU-MMD | accuracy: 65.6% vs 65.8%; 79.2% vs 79.5%; 81.6% vs 82.9%; 30.0% vs 31.4%; 68.5% vs 69.3%; 78.6% vs 79.4% | No clear effect |
| Use the top convolutional layer for distance regression instead of the selected layer (ablation) | SSNet · OAD | SL-Score: 0.61 -> 0.69 | Improved |
| Use the top convolutional layer for distance regression instead of the selected layer (ablation) | SSNet · PKU-MMD | SL-Score: 0.67 -> 0.72 | Improved |
| Increase the convolutional depth from 8 to 14 layers (ablation) | SSNet · OAD | prediction accuracy: 73.4 -> 78.7 | Improved |
| Increase the convolutional depth from 14 to 16 layers (ablation) | SSNet · OAD | prediction accuracy: 78.7; 78.7 | No clear effect |
| Vary the regression-loss weight gamma (ablation) | SSNet · OAD | accuracy: r = 0.001; r = 0.01; r = 0.05; r = 0.1; r = 0.3 | Mixed |
Closest papers by meaning
- Skeleton-Based Online Action Prediction Using Scale Selection Network (TPAMI 2019 · similarity 0.96)
- Online Human Action Detection Using Joint Classification-Regression Recurrent Neural Networks (ECCV 2016 · similarity 0.90)
- Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks (CVPR 2017 · similarity 0.87)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.87)
- A New Representation of Skeleton Sequences for 3D Action Recognition (CVPR 2017 · similarity 0.86)
- Spatio-Temporal Naive-Bayes Nearest-Neighbor (ST-NBNN) for Skeleton-Based Action Recognition (CVPR 2017 · 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.
