Research map / 3D vision / Face and shape
ARAH: Animatable Volume Rendering of Articulated Human SDFs
Shaofei Wang, Katja Schwarz, Andreas Geiger, Siyu Tang
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 2 extra root-finding initializations from nearest SMPL bones (ablation) | ARAH · ZJU-MoCap seq 313 novel views | PSNR/SSIM/LPIPS: 31.6/0.973/0.050 -> 31.5/0.972/0.049 | No clear effect |
| use joint root-finding with learned forward LBS instead of backward LBS network (Ani-NeRF) | ARAH · ZJU-MoCap seq 313 unseen poses | LPIPS: 0.115 -> 0.092 | Improved |
| use ARAH instead of Neural Body on unseen poses | ARAH · ZJU-MoCap seq 315 unseen poses | LPIPS: 0.152 -> 0.105 | Improved |
| model geometry as articulated SDF instead of Neural Body NeRF | ARAH · ZJU-MoCap seq 313 geometry | Chamfer distance: 1.258 -> 0.707 | Improved |
| model geometry as articulated SDF instead of Neural Body NeRF | ARAH · ZJU-MoCap seq 313 geometry | normal consistency: 0.700 -> 0.809 | Improved |
| use ARAH instead of Neural Body for novel view synthesis of training poses | ARAH · ZJU-MoCap seq 377 training poses | PSNR: 28.1 -> 27.8 | Mixed |
| use ARAH instead of Neural Body on novel views of training poses | ARAH · ZJU-MoCap seq 313 training poses | LPIPS: 0.068 -> 0.050 | Improved |
| use ARAH instead of NARF on H36M unseen poses (subject S5) | ARAH · H36M unseen poses S5 | PSNR: 23.91 -> 22.79 | Hurt |
| use ARAH instead of Ani-N on H36M unseen poses | ARAH · H36M unseen poses (average) | PSNR: 22.55 -> 23.42 | Improved |
| use the proposed ray sampling instead of surface rendering or uniform sampling with 64 samples (ablation) | ARAH · ZJU-MoCap seq 313 | visual geometry artifacts | Improved |
| use learned forward LBS instead of backward LBS network (ablation) | ARAH · ZJU-MoCap seq 313 | visual artifacts | Improved |
| use learned forward LBS instead of deterministic nearest-neighbor SMPL skinning weights (ablation) | ARAH · ZJU-MoCap seq 313 | visual artifacts | Improved |
Later papers that compared with this method
7 later accepted papers report ARAH as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.
- 3D²-Actor: Learning Pose-Conditioned 3D-Aware Denoiser for Realistic Gaussian Avatar Modeling (AAAI 2025 · better on 9 of 13 shared results)
- MetaCap: Meta-learning Priors from Multi-View Imagery for Sparse-view Human Performance Capture and Rendering (ECCV 2024 · better on 10 of 11 shared results)
- 3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting (CVPR 2024 · better on 6 of 6 shared results)
- NECA: Neural Customizable Human Avatar (CVPR 2024 · better on 3 of 4 shared results)
- Relightable and Animatable Neural Avatars from Videos (AAAI 2024 · better on 2 of 2 shared results)
- MeshAvatar: Learning High-quality Triangular Human Avatars from Multi-view Videos (ECCV 2024 · better on 1 of 1 shared result)
- NPC: Neural Point Characters from Video (ICCV 2023 · better on 7 of 9 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Neural Body — Neural Body: Implicit Neural Representations With Structured Latent Codes for Novel View Synthesis of Dynamic Humans (CVPR 2021) · better on 15 of 15 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
- 3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting (CVPR 2024 · similarity 0.90)
- AniGS: Animatable Gaussian Avatar from a Single Image with Inconsistent Gaussian Reconstruction (CVPR 2025 · similarity 0.90)
- Relightable and Animatable Neural Avatar from Sparse-View Video (CVPR 2024 · similarity 0.89)
- Human Gaussian Splatting: Real-time Rendering of Animatable Avatars (CVPR 2024 · similarity 0.89)
- IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing (CVPR 2024 · similarity 0.88)
- Relightable and Animatable Neural Avatars from Videos (AAAI 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.
