Research map / 3D vision / NeRF and radiance fields
Learning Neural Volumetric Representations of Dynamic Humans in Minutes
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a part-based body representation instead of one shared NeRF (ablation) | dynamic human representation · ZJU-MoCap 377 | PSNR: 30.11 -> 32.09 | Helped |
| use a part-based body representation instead of one shared NeRF (ablation) | dynamic human representation · ZJU-MoCap 377 | LPIPS*: 45.84 -> 23.47 | Helped |
| parameterize residual motion with surface coordinates (u, v, t) instead of hash-encoded (x, t) (ablation) | dynamic human representation · ZJU-MoCap 377 | PSNR: 31.40 -> 32.09 | Helped |
| parameterize residual motion with surface coordinates (u, v, t) instead of hash-encoded (x, t) (ablation) | dynamic human representation · ZJU-MoCap 377 | LPIPS*: 30.99 -> 23.47 | Helped |
| add perceptual loss during training (ablation) | dynamic human representation · ZJU-MoCap 377 | LPIPS*: 44.33 -> 23.47 | Helped |
| use hash-encoded (u, v, t) rather than positional-encoded (x, t) as the residual-motion input (ablation) | dynamic human representation · ZJU-MoCap 377 | PSNR: 31.94 -> 32.09 | Helped |
| use hash-encoded (u, v, t) rather than hash-encoded x with a per-frame latent code (ablation) | dynamic human representation · ZJU-MoCap 377 | PSNR: 31.32 -> 32.09 | Helped |
| use hash-encoded (u, v, t) rather than hash-encoded x with pose parameters (ablation) | dynamic human representation · ZJU-MoCap 377 | PSNR: 31.51 -> 32.09 | Helped |
| allocate hash-table sizes by human part instead of using a uniform size of 2^15 (ablation) | part-based voxelized human representation · ZJU-MoCap 377 | PSNR: 30.69 -> 32.09 | Helped |
| allocate hash-table sizes by human part instead of using a uniform size of 2^20 (ablation) | part-based voxelized human representation · ZJU-MoCap 377 | PSNR: 31.18 -> 32.09 | Helped |
Closest papers by meaning
- H-NeRF: Neural Radiance Fields for Rendering and Temporal Reconstruction of Humans in Motion (NeurIPS 2021 · similarity 0.91)
- HumanNeRF: Efficiently Generated Human Radiance Field From Sparse Inputs (CVPR 2022 · similarity 0.90)
- PKU-DyMVHumans: A Multi-View Video Benchmark for High-Fidelity Dynamic Human Modeling (CVPR 2024 · similarity 0.88)
- Relightable and Animatable Neural Avatar from Sparse-View Video (CVPR 2024 · similarity 0.87)
- SurMo: Surface-based 4D Motion Modeling for Dynamic Human Rendering (CVPR 2024 · similarity 0.87)
- DynaVol: Unsupervised Learning for Dynamic Scenes through Object-Centric Voxelization (ICLR 2024 · 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.