Research map / 3D vision / Face and shape
NASA Neural Articulated Shape Approximation
Boyang Deng, JP Lewis, Timothy Jeruzalski, Gerard Pons-Moll, Geoffrey Hinton, Mohammad Norouzi, Andrea Tagliasacchi
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 |
|---|---|---|---|
| decompose into per-bone rigid parts (rigid R vs unstructured U) (ablation) | NASA · AMASS DFaust | F-score: 46.15 -> 93.94 (+49% F-score) | Improved |
| add pose-dependent deformation of parts (D vs R) (ablation) | NASA · AMASS DFaust | F-score: 93.94 -> 98.54 (+5% F-score) | Improved |
| decompose into per-bone rigid parts (R vs U) (ablation) | NASA · AMASS Transitions | F-score: 26.83 -> 96.71 (+70% F-score) | Improved |
| add deformable parts (D vs R) (ablation) | NASA · AMASS Transitions | mIoU: .936 -> .965 | Improved |
| train on more diverse data (Transitions instead of DFaust) | NASA-D · AMASS | F-score: 98.54 -> 99.42 (+1% F-score) | Improved |
| use structured (D) instead of unstructured (U) representation for tracking (ablation) | NASA tracking · DFaust easy (00-01) | F-score: 61.63 -> 99.08 | Improved |
| use structured (D) instead of unstructured (U) representation for tracking (ablation) | NASA tracking · DFaust hard (02-09) | F-score: 50.63 -> 96.48 | Improved |
| use the ground-truth (oracle) pose instead of the tracked pose (ablation) | NASA-D · DFaust easy (00-01) | F-score: 99.08 -> 99.01 | No clear effect |
| use the ground-truth (oracle) pose instead of the tracked pose (ablation) | NASA-D · DFaust hard (02-09) | F-score: 96.48 -> 96.80 | No clear effect |
| use the ground-truth (oracle) pose instead of the tracked pose (ablation) | NASA-D · DFaust easy (00-01) | mIoU: .968 -> .976 | Improved |
| use the ground-truth (oracle) pose instead of the tracked pose (ablation) | NASA-D · DFaust hard (02-09) | mIoU: .948 -> .959 | Improved |
Closest papers by meaning
- BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022 · similarity 0.86)
- NPMs: Neural Parametric Models for 3D Deformable Shapes (ICCV 2021 · similarity 0.86)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.85)
- Neural Shape Deformation Priors (NeurIPS 2022 · similarity 0.85)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.85)
- CAGE: Controllable Articulation GEneration (CVPR 2024 · 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.
