Research map / 3D vision / Face and shape
NPC: Neural Point Characters from Video
Shih-Yang Su, Timur Bagautdinov, Helge Rhodin
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 |
|---|---|---|---|
| use NPC point-based representation instead of template-free baselines (DANBO, TAVA) | NPC · Human3.6M novel-view (avg) | LPIPS: 0.115 -> 0.104 | Improved |
| NPC vs DANBO on unseen pose synthesis | NPC · MonoPerfCap (avg) | KIDx100: 4.74 -> 3.07 | Improved |
| NPC vs DANBO on MonoPerfCap LPIPS | NPC · MonoPerfCap (avg) | LPIPS: 0.202 -> 0.202 | No clear effect |
| NPC vs scan-prior ARAH on ZJU-Mocap pixel-wise PSNR | NPC · ZJU-Mocap subject 387 | PSNR: 25.83 -> 25.10 (novel view); 22.93 -> 21.88 (novel pose) | Hurt |
| NPC vs ARAH on ZJU-Mocap FID | NPC · ZJU-Mocap subject 387 | FID: 38.2 -> 36.1 (novel view); 51.4 -> 49.2 (novel pose) | Improved |
| add coordinate feature f^d (ablation) | NPC · Human3.6M S9 | KIDx100: 5.54 -> 4.43 | Improved |
| add per-point feature f^p (ablation) | NPC · Human3.6M S9 | KIDx100: 4.51 -> 4.43 | Improved |
| add pose-dependent feature f^theta (ablation) | NPC · Human3.6M S9 | KIDx100: 4.50 -> 4.43 | Improved |
| add non-linear point deformation Delta p (ablation) | NPC · Human3.6M S9 | KIDx100: 4.61 -> 4.43 | Improved |
| increase points per body part beyond 200 (ablation) | NPC · Human3.6M | LPIPS: 0.115 -> 0.114 (280x19) | No clear effect |
| re-run training with 4 different canonical point initializations (ablation) | NPC · Human3.6M S9 | KIDx100: 4.34 (± 0.12) | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ARAH — ARAH: Animatable Volume Rendering of Articulated Human SDFs (ECCV 2022) · better on 7 of 9 shared results
- DANBO — DANBO: Disentangled Articulated Neural Body Representations via Graph Neural Networks (ECCV 2022) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022 · similarity 0.89)
- Hierarchically Structured Neural Bones for Reconstructing Animatable Objects from Casual Videos (ECCV 2024 · similarity 0.89)
- Animatable Implicit Neural Representations for Creating Realistic Avatars From Videos (TPAMI 2024 · similarity 0.88)
- NPMs: Neural Parametric Models for 3D Deformable Shapes (ICCV 2021 · similarity 0.88)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.87)
- Learning Implicit Representation for Reconstructing Articulated Objects (ICLR 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.
