Research map / Video / Action recognition
Skeleton-Based Action Recognition Using Spatio-Temporal LSTM Network with Trust Gates
Jun Liu, Amir Shahroudy, Dong Xu, Alex Chichung Kot, Gang Wang
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 |
|---|---|---|---|
| use tree traversal instead of joint chain (ablation) | ST-LSTM · NTU RGB+D (X-Subject) | accuracy: 61.7% -> 65.2% | Improved |
| add trust gate to tree ST-LSTM (ablation) | ST-LSTM (Tree) · NTU RGB+D (X-Subject) | accuracy: 65.2% -> 69.2% | Improved |
| add trust gate to tree ST-LSTM (ablation) | ST-LSTM (Tree) · NTU RGB+D side views | accuracy: 76.5% -> 81.6% | Improved |
| add trust gate to tree ST-LSTM (ablation) | ST-LSTM (Tree) · SBU Interaction | accuracy: 88.6% -> 93.3% | Improved |
| add trust gate to tree ST-LSTM (ablation) | ST-LSTM (Tree) · UT-Kinect (LOOCV) | accuracy: 92.4% -> 97.0% | Improved |
| add trust gate to tree ST-LSTM (ablation) | ST-LSTM (Tree) · SYSU-3D | accuracy: 73.4% -> 76.5% | Improved |
| stack two ST-LSTM layers instead of one (ablation) | ST-LSTM (Tree) + Trust Gate · NTU RGB+D (X-Subject) | accuracy: 65.5% -> 69.2% | Improved |
| fuse visual features inside the ST-LSTM unit versus early fusion (ablation) | ST-LSTM + Trust Gate · NTU RGB+D (X-Subject) | accuracy: 70.8% -> 73.2% | Improved |
| add visual features by in-unit fusion (ablation) | ST-LSTM + Trust Gate · NTU RGB+D (X-Subject) | accuracy: 69.2% -> 73.2% | Improved |
| add last-to-first link (ablation) | ST-LSTM (Tree) + Trust Gate · NTU RGB+D (X-Subject) | accuracy: 68.5% -> 69.2% | Improved |
| add skeleton rotation normalization (ablation) | ST-LSTM (Tree) + Trust Gate · SYSU-3D | accuracy: 73.0% -> 76.5% | Improved |
| use tree traversal instead of double joint chain (ablation) | ST-LSTM · SBU Interaction | accuracy: 85.9% -> 88.6% | Improved |
Closest papers by meaning
- Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition (ECCV 2016 · similarity 0.94)
- Co-Occurrence Feature Learning for Skeleton Based Action Recognition Using Regularized Deep LSTM Networks (AAAI 2016 · similarity 0.93)
- Ensemble Deep Learning for Skeleton-Based Action Recognition Using Temporal Sliding LSTM Networks (ICCV 2017 · similarity 0.92)
- An Attention Enhanced Graph Convolutional LSTM Network for Skeleton-Based Action Recognition (CVPR 2019 · similarity 0.92)
- An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Data (AAAI 2017 · similarity 0.90)
- Bayesian Graph Convolution LSTM for Skeleton Based Action Recognition (ICCV 2019 · similarity 0.90)
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.
