Research map / 3D vision / Face and shape
DSFNet: Dual Space Fusion Network for Occlusion-Robust 3D Dense Face Alignment
Read the paperCheck 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 |
|---|---|---|---|
| Fuse image-space and model-space predictions using the fusion module | DSFNet · AFLW2000-3D, 3D dense alignment | NME (%): 3.89 -> 3.80 | Helped |
| Add the model-space feature f_ms to the fusion module (ablation) | DSFNet-f · AFLW2000-3D, 3D dense alignment | NME (%): 3.96 -> 3.80 | Helped |
| Add the model-space feature f_ms to the fusion module (ablation) | DSFNet-f · AFLW2000-3D-occlusion NOD | NME (%) | No clear effect |
| Add correspondence-map UV coordinates as PointNet input (ablation) | Image-space post-process · AFLW2000-3D | NME: 4.01 -> 3.93 | Helped |
| Add confidence-map values as PointNet input (ablation) | Image-space post-process · AFLW2000-3D | NME: 3.93 -> 3.94 | No clear effect |
| Add the geometry-consistency loss Lcons (ablation) | Image-space post-process · AFLW2000-3D | NME: 3.94 -> 3.89 | Helped |
| Align the point cloud to canonical view before PointNet (ablation) | Image-space post-process · AFLW2000-3D | NME: 4.12 -> 3.89 | Helped |
| Combine image-space and model-space predictions | DSFNet-f · AFLW2000-3D-occlusion CSOD | NME (%): 4.59 -> 4.44 | Helped |
| Combine image-space and model-space predictions | DSFNet-f · AFLW2000-3D-occlusion NSOD | NME (%): 5.83 -> 5.40 | Helped |
| Combine image-space and model-space predictions | DSFNet-f · AFLW2000-3D-occlusion NOD | NME (%): 4.78 -> 4.80 | Hurt |
Closest papers by meaning
- Face Alignment Across Large Poses: A 3D Solution (CVPR 2016 · similarity 0.89)
- MVF-Net: Multi-View 3D Face Morphable Model Regression (CVPR 2019 · similarity 0.88)
- Regressing Robust and Discriminative 3D Morphable Models With a Very Deep Neural Network (CVPR 2017 · similarity 0.87)
- Large-Pose Face Alignment via CNN-Based Dense 3D Model Fitting (CVPR 2016 · similarity 0.87)
- MMFace: A Multi-Metric Regression Network for Unconstrained Face Reconstruction (CVPR 2019 · similarity 0.87)
- Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction (ICLR 2026 · similarity 0.87)
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.