Research map / 3D vision / Face and shape
NPMs: Neural Parametric Models for 3D Deformable Shapes
Pablo Palafox, Aljaž Božič, Justus Thies, Matthias Nießner, Angela Dai
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 |
|---|---|---|---|
| initialize pose code with the pose encoder (ablation) | NPMs · CAPE | EPE: 3.61 -> 0.74 | Improved |
| initialize shape code with the shape encoder (ablation) | NPMs · CAPE | EPE: 0.77 -> 0.74 | No clear effect |
| initialize both codes with encoders (ablation) | NPMs · CAPE | IoU: 0.77 -> 0.83 | Improved |
| add positional encoding to shape and pose MLPs (ablation) | NPMs · CAPE | Chamfer-l2: 0.142 -> 0.022 | Improved |
| use smaller latent code size 128 (ablation) | NPMs · CAPE | Chamfer-l2: 0.052 -> 0.022 | Hurt |
| use larger latent code size 384 (ablation) | NPMs · CAPE | Chamfer-l2: 0.076 -> 0.022 | Hurt |
| add temporal consistency on flow predictions (ablation) | NPMs · CAPE | Chamfer-l2: 0.037 -> 0.022 | Improved |
| enforce temporal consistency on flow instead of on pose codes (ablation) | NPMs · CAPE | Chamfer-l2: 0.034 -> 0.022 | Improved |
| add ICP loss at the start of optimization (ablation) | NPMs · CAPE | Chamfer-l2: 0.031 -> 0.022 | Improved |
| add ICP loss at the start of optimization (ablation) | NPMs · CAPE | EPE: 0.79 -> 0.74 | Improved |
Closest papers by meaning
- SPAMs: Structured Implicit Parametric Models (CVPR 2022 · similarity 0.91)
- BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022 · similarity 0.90)
- Unsupervised Shape and Pose Disentanglement for 3D Meshes (ECCV 2020 · similarity 0.89)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.89)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.89)
- Neural Parametric Gaussians for Monocular Non-Rigid Object Reconstruction (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.
