Research map / 3D vision / Face and shape
Unsupervised Learning of Efficient Geometry-Aware Neural Articulated Representations
Atsuhiro Noguchi, Xiao Sun, Stephen Lin, Tatsuya Harada
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 tri-plane representation to NARF (ENARF) | NARF · ZJU mocap dynamic scenes | render time (s): 2.17 -> 0.47 | Improved |
| add tri-plane representation to NARF (ENARF) | NARF · ZJU mocap dynamic scenes (novel view) | PSNR: 30.62 -> 31.94 | Improved |
| add tri-plane deformation fields (D-ENARF) | ENARF · ZJU mocap dynamic scenes (novel view) | PSNR: 31.94 -> 32.93 | Improved |
| add tri-plane deformation fields (D-ENARF) | ENARF · ZJU mocap dynamic scenes (novel pose) | SSIM: 0.953 -> 0.9396 | Hurt |
| add tri-plane selector (ablation) | ENARF · ZJU mocap dynamic scenes (novel view) | PSNR: 29.16 -> 31.94 | Improved |
| replace tri-plane selector with MLP selector (ablation) | ENARF · ZJU mocap dynamic scenes | render time (s): 0.47 -> 1.13 | Hurt |
| replace tri-plane selector with MLP selector (ablation) | ENARF · ZJU mocap dynamic scenes (novel view) | PSNR: 31.94 -> 32.27 | Mixed |
| add view direction input to decoder (ablation) | ENARF · ZJU mocap dynamic scenes (novel view) | PSNR: 31.94 -> 29.98 | Hurt |
| add bone region loss (ablation) | ENARF-GAN · SURREAL | PCKh@0.5: 0.863 -> 0.947 | Improved |
| add bone region loss (ablation) | ENARF-GAN · SURREAL | Depth: 14.8 -> 8.8 | Improved |
| add bone region loss (ablation) | ENARF-GAN · SURREAL | FG-FID: 21.0 -> 21.3 | No clear effect |
| use CMU Panoptic pose prior instead of ground-truth dataset poses (ablation) | ENARF-GAN · SURREAL | FID: 22.6 -> 24.2 | Hurt |
Closest papers by meaning
- HoloGAN: Unsupervised Learning of 3D Representations From Natural Images (ICCV 2019 · similarity 0.91)
- Intrinsic-Extrinsic Preserved GANs for Unsupervised 3D Pose Transfer (ICCV 2021 · similarity 0.88)
- Neural Articulated Radiance Field (ICCV 2021 · similarity 0.88)
- Unsupervised Geometry-Aware Representation for 3D Human Pose Estimation (ECCV 2018 · similarity 0.88)
- Self-Supervised Neural Articulated Shape and Appearance Models (CVPR 2022 · similarity 0.87)
- Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis (CVPR 2020 · 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.
