Research map / Video / Action recognition
Sparse Coding of Shape Trajectories for Facial Expression and 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 |
|---|---|---|---|
| project skeletons into Kendall shape space and code with Kendall SCDL (ablation) | raw normalized data · NTU-RGB+D | accuracy (%): 56.5 -> 82.95 | Improved |
| project skeletons into Kendall shape space and code with Kendall SCDL (ablation) | raw normalized data · Florence 3D | accuracy (%): 84.29 -> 92.85 | Improved |
| use nonlinear (Kendall) SCDL instead of linear SCDL (ablation) | linear SCDL · NTU-RGB+D | accuracy (%): 79.20 -> 82.95 | Improved |
| use nonlinear (Kendall) SCDL instead of linear SCDL (ablation) | linear SCDL · UTKinect | accuracy (%): 93.58 -> 97.5 | Improved |
| use nonlinear (Kendall) SCDL instead of linear SCDL (ablation) | linear SCDL · MSR 3D | accuracy (%): 89.23 -> 90.01 | Improved |
| use Chehra 49 landmarks instead of OpenFace 49 landmarks (ablation) | OpenFace · CK+ | accuracy (%): 83.73 -> 93.68 | Improved |
| use OpenFace 68 landmarks instead of 49 (ablation) | OpenFace · Oulu-CASIA | accuracy (%): 70.85 -> 71.26 | No clear effect |
| use extrinsic instead of intrinsic SCDL for 2D expressions | intrinsic SCDL (SVM) · CK+ | accuracy (%): 91.26 -> 95.62 | Improved |
| use extrinsic instead of intrinsic SCDL for micro-expressions | intrinsic SCDL (SVM) · CASME II | accuracy (%): 43.65 -> 64.62 | Improved |
| use intrinsic instead of extrinsic SCDL for 3D actions | extrinsic SCDL (SVM) · MSR-Action 3D (Half-Half) | accuracy (%): 82.52 -> 90.01 | Improved |
| use Bi-LSTM instead of SVM with extrinsic SCDL | extrinsic SCDL · Oulu-CASIA | accuracy (%): 77.06 -> 73.09 | Hurt |
| use Bi-LSTM instead of DTW-FTP-SVM with intrinsic SCDL | intrinsic SCDL · MSR-Action 3D (3 Subsets) | accuracy (%): 94.19 -> 86.18 | Hurt |
Closest papers by meaning
- Coding Kendall's Shape Trajectories for 3D Action Recognition (CVPR 2018 · similarity 0.92)
- A Novel Geometric Framework on Gram Matrix Trajectories for Human Behavior Understanding (TPAMI 2018 · similarity 0.90)
- Action Recognition Using Rate-Invariant Analysis of Skeletal Shape Trajectories (TPAMI 2016 · similarity 0.89)
- A Novel Space-Time Representation on the Positive Semidefinite Cone for Facial Expression Recognition (ICCV 2017 · similarity 0.89)
- Elastic Functional Coding of Riemannian Trajectories (TPAMI 2016 · similarity 0.87)
- TrAction: Action Recognition with Sparse Trajectories (NeurIPS 2026 · 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.
