Research map / 3D vision / Face and shape
Coding Kendall's Shape Trajectories for 3D Action Recognition
Amor Ben Tanfous, Hassen Drira, Boulbaba Ben Amor
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 class-specific dictionaries instead of a global dictionary (ablation) | Kendall sparse coding with Bi-LSTM · Florence3D-Action, LOAO | accuracy: 91.53% -> 94.48% | Improved |
| use a global dictionary instead of class-specific dictionaries (ablation) | Kendall sparse coding with Bi-LSTM · Florence3D-Action, LOAO | accuracy: 94.48% -> 91.53% | Hurt |
| use bidirectional LSTM instead of traditional LSTM (ablation) | Kendall sparse-code trajectories · Florence3D-Action, LOAO | accuracy: around 0.7% | Improved |
| initialize the dictionary with Bayesian shape clustering and PGA rather than random training shapes (ablation) | Kendall dictionary learning · Florence3D-Action, tight lace class | dictionary learning convergence speed: approximately two times | Improved |
| increase sparsity regularization to λ = 0.01 (ablation) | Kendall sparse coding with FTP-SVM · Florence3D-Action validation split | accuracy: 37% of sparsity | Improved |
| increase sparsity regularization to λ = 0.02 (ablation) | Kendall sparse coding with Bi-LSTM · Florence3D-Action validation split | accuracy: 49% of sparsity | Improved |
| use Kendall sparse-code functions with FTP-SVM instead of the compared Riemannian representation | 3D action recognition · MSR-Action 3D, protocol [28] | accuracy: 92.46 -> 94.19 | Improved |
| use Kendall sparse-code functions with FTP-SVM instead of the compared Riemannian representation | 3D action recognition · Florence3D-Action, protocol [38] | accuracy: 90.8 -> 92.85 | Improved |
| use Kendall sparse-code functions with Bi-LSTM instead of the compared Riemannian representation | 3D action recognition · UTKinect-Action, protocol [38] | accuracy: 97.08 -> 96.89 | Hurt |
| use Kendall sparse-code functions with FTP-SVM instead of the compared Riemannian representation | 3D action recognition · UTKinect-Action, protocol [38] | accuracy: 97.08 -> 97.39 | Improved |
| use Bi-LSTM without an additional noise-processing step instead of FTP-SVM | Kendall sparse-code trajectories · MSR-Action 3D, protocol [28] | accuracy: 94.19 -> 91.84 | Hurt |
Closest papers by meaning
- Sparse Coding of Shape Trajectories for Facial Expression and Action Recognition (TPAMI 2019 · similarity 0.92)
- Elastic Functional Coding of Riemannian Trajectories (TPAMI 2016 · similarity 0.89)
- Action Recognition Using Rate-Invariant Analysis of Skeletal Shape Trajectories (TPAMI 2016 · similarity 0.89)
- Efficient Temporal Sequence Comparison and Classification Using Gram Matrix Embeddings on a Riemannian Manifold (CVPR 2016 · similarity 0.86)
- TrAction: Action Recognition with Sparse Trajectories (NeurIPS 2026 · similarity 0.84)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.84)
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.
