Research map / 3D vision / Face and shape
Fast-SNARF: A Fast Deformer for Articulated Neural Fields
Xu Chen, Tianjian Jiang, Jie Song, Max Rietmann, Andreas Geiger, Michael J. Black, Otmar Hilliges
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 voxel-based correspondence search (ablation) | SNARF · articulation inference time | time: 807ms -> 61ms | Improved |
| add pre-computed LBS (ablation) | Fast-SNARF · articulation inference time | time: 61ms -> 40ms | Improved |
| add custom CUDA kernel (ablation) | Fast-SNARF · articulation inference time | time: 40ms -> 5.3ms | Improved |
| add filtering of duplicate correspondences (ablation) | Fast-SNARF · shape query time | time: 187ms -> 28ms | Improved |
| add MLP distillation to voxel skinning weights (ablation) | Fast-SNARF · PosePrior OOD surface IoU | accuracy: 78.2% -> 81.2% | Improved |
| add MLP distillation to voxel skinning weights (ablation) | Fast-SNARF · training time | time: 23 min -> 25 min | Hurt |
| raise voxel resolution from 16x16x4 to 64x64x16 (ablation) | Fast-SNARF · PosePrior OOD surface IoU | accuracy: 78.3% -> 81.2% | Improved |
| raise voxel resolution from 64x64x16 to 128x128x32 (ablation) | Fast-SNARF · PosePrior OOD surface IoU | accuracy: 81.2% -> 81.5% | No clear effect |
| raise voxel resolution from 64x64x16 to 128x128x32 (ablation) | Fast-SNARF · training time | time: 25min -> 52min | Hurt |
| replace SNARF with Fast-SNARF (ablation) | SNARF · DFaust training time | time: 8h -> 25min | Improved |
| replace SNARF with Fast-SNARF | SNARF · PosePrior OOD surface IoU | accuracy: 80.65% -> 81.25% | No clear effect |
Closest papers by meaning
- SNARF: Differentiable Forward Skinning for Animating Non-Rigid Neural Implicit Shapes (ICCV 2021 · similarity 0.93)
- Animatable Implicit Neural Representations for Creating Realistic Avatars From Videos (TPAMI 2024 · similarity 0.88)
- BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022 · similarity 0.87)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.87)
- NRDF: Neural Riemannian Distance Fields for Learning Articulated Pose Priors (CVPR 2024 · similarity 0.86)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.86)
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.
