Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Hierarchical Dynamic Model for Human Action Recognition
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 |
|---|---|---|---|
| Use hierarchical parameters to model within-class variation, rather than non-hierarchical HMM/HSMM | HDM · MSRA, UTD, G3D, Penn | accuracy (%): 67.8 -> 86.1; 82.8 -> 92.8; 68.1 -> 92.0; 82.3 -> 93.4 | Helped |
| Learn hyperparameters instead of using initial values (ablation) | HDM · MSRA | accuracy (%): 70.3 -> 80.6; 6.5% | Helped |
| Use Bayesian variational inference rather than point-estimate predictive likelihood (ablation) | HDM · MSRA | accuracy (%): 80.6 -> 82.1; 0.5% | Helped |
| Use Gibbs sampling rather than point-estimate predictive likelihood (ablation) | HDM · MSRA | accuracy (%): 80.6 -> 86.1; 3.3% | Helped |
| Use HDM rather than the strongest listed state-of-the-art baseline | HDM · UTD | accuracy (%): 87.9 -> 92.8 | Helped |
| Use HDM rather than the strongest listed state-of-the-art baseline | HDM · Penn | accuracy (%): 93.2 -> 93.4 | Helped |
| Use HDM rather than the strongest listed state-of-the-art baseline | HDM · MSRA | accuracy (%): 92.0 -> 86.1 | Hurt |
| Use HDM rather than the strongest listed state-of-the-art baseline | HDM · G3D | accuracy (%): 96.0 -> 92.0 | Hurt |
| Train on combined datasets for cross-dataset evaluation rather than HSMM | HDM · Cross-dataset, pairwise training and individual dataset test | accuracy (%): 73.7 -> 86.9; 82.5 -> 91.9; 89.0 -> 93.6; 87.0 -> 92.6; 91.0 -> 97.3; 83.2 -> 91.0 | Helped |
| Train on two datasets and test on the remaining dataset rather than HSMM | HDM · Cross-dataset, two datasets train and third dataset test | accuracy (%): 65.3 -> 89.2; 61.9 -> 75.0; 42.5 -> 61.2 | Helped |
| Handle randomly missing observations using the generative likelihood | HDM · MSRA, UTD, G3D and Penn with missing data | accuracy (%) | Helped |
| Use HDM rather than LSTM across individual datasets | HDM · MSRA, UTD, G3D, Penn | accuracy (%): 74.7 -> 86.1; 77.0 -> 92.8; 82.2 -> 92.0; 90.3 -> 93.4 | Mixed |
Closest papers by meaning
- A Hierarchical Pose-Based Approach to Complex Action Understanding Using Dictionaries of Actionlets and Motion Poselets (CVPR 2016 · similarity 0.86)
- Uncertainty-Guided Probabilistic Transformer for Complex Action Recognition (CVPR 2022 · similarity 0.85)
- Stochastic Human Motion Prediction with Memory of Action Transition and Action Characteristic (CVPR 2025 · similarity 0.83)
- VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention (AAAI 2025 · similarity 0.83)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.83)
- Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition (ICML 2024 · similarity 0.83)
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.