Research map / 3D vision / Face and shape
Faster Than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses
Chandrasekhar Bhagavatula, Chenchen Zhu, Khoa Luu, Marios Savvides
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 |
|---|---|---|---|
| add landmark regression stage to AlexNet model (ablation) | 3DSTN (AlexNet) · AFLW detected (21 pts) | mean NME (%): 5.84 -> 4.79 | Improved |
| add landmark regression stage to VGG-16 model (ablation) | 3DSTN (VGG-16) · AFLW detected (21 pts) | mean NME (%): 4.92 -> 4.00 | Improved |
| use VGG-16 instead of AlexNet pretrained backbone (ablation) | 3DSTN · AFLW detected (21 pts) | mean NME (%): 5.84 -> 4.92 | Improved |
| use VGG-16 instead of AlexNet backbone, with landmark regression | 3DSTN · AFLW (21 pts) comparison | mean NME (%): 5.14 -> 4.23 | Improved |
| 3DSTN VGG-16 vs 3DDFA+SDM | 3DSTN (VGG-16) · AFLW (21 pts) | mean NME (%): 5.32 -> 4.23 | Improved |
| 3DSTN VGG-16 vs 3DDFA+SDM | 3DSTN (VGG-16) · AFLW2000-3D (68 pts) | mean NME (%): 4.94 -> 4.49 | Improved |
| 3DSTN AlexNet vs 3DDFA+SDM | 3DSTN (AlexNet) · AFLW2000-3D (68 pts) | mean NME (%): 4.94 -> 5.41 | Hurt |
| 3DSTN VGG-16 vs 3DDFA+SDM | 3DSTN (VGG-16) · AFLW2000-3D (30-60 deg) | NME (%): 4.24 -> 4.33 | Hurt |
| 3DSTN AlexNet vs 3DDFA+SDM | 3DSTN (AlexNet) · AFLW (21 pts) | mean NME (%): 5.32 -> 5.14 | Improved |
| 3DSTN VGG-16 vs 3DDFA+SDM | 3DSTN (VGG-16) · AFLW (21 pts) | NME std (%): 0.92 -> 0.87 | Improved |
| single-pass 3DSTN AlexNet vs iterative 3DDFA | 3DSTN (AlexNet) · AFLW speed | time per image: 75.72 ms -> 0.0059 seconds | Improved |
| 3DSTN VGG-16 vs 3DDFA | 3DSTN (VGG-16) · AFLW speed | time per image: 75.72 ms -> 0.0190 seconds | Improved |
Closest papers by meaning
- How Far Are We From Solving the 2D & 3D Face Alignment Problem? (And a Dataset of 230,000 3D Facial Landmarks) (ICCV 2017 · similarity 0.93)
- Face Alignment Across Large Poses: A 3D Solution (CVPR 2016 · similarity 0.90)
- Large Pose 3D Face Reconstruction From a Single Image via Direct Volumetric CNN Regression (ICCV 2017 · similarity 0.90)
- Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network (ECCV 2018 · similarity 0.90)
- Face Alignment in Full Pose Range: A 3D Total Solution (TPAMI 2017 · similarity 0.89)
- FaceLift: Semi-supervised 3D Facial Landmark Localization (CVPR 2024 · similarity 0.88)
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.
