Research map / 3D vision / 3D generation
AvatarStudio: High-Fidelity and Animatable 3D Avatar Creation from Text
Xuanmeng Zhang, Jianfeng Zhang, Chenxu Zhang, Jun Hao Liew, Huichao Zhang, Yi Yang, Jiashi Feng
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 DensePose-conditioned ControlNet instead of Stable Diffusion guidance (ablation) | AvatarStudio · pose controllability | SMPL MSE (x10^-2): 9.0 -> 5.9 | Improved |
| use DensePose-conditioned ControlNet instead of skeleton ControlNet (ablation) | AvatarStudio · pose controllability | SMPL MSE (x10^-2): 7.7 -> 5.9 | Improved |
| add dual-space (canonical and deformed) training (ablation) | AvatarStudio · pose controllability | SMPL MSE (x10^-2): 6.5 -> 5.9 | Improved |
| combine NeRF and mesh instead of mesh-only (DMTet) (ablation) | AvatarStudio · avatar quality | qualitative | Improved |
| combine NeRF and mesh instead of NeRF-only (ablation) | AvatarStudio · avatar quality | qualitative | Improved |
| add CFG rescale (ablation) | AvatarStudio · avatar appearance | qualitative | Improved |
| add part-aware super-resolution (ablation) | AvatarStudio · avatar appearance | qualitative | Improved |
| use AvatarStudio instead of DreamWaltz | text-to-3D avatar · specific characters | CLIP score: 31.52 -> 32.80 | Improved |
| use AvatarStudio instead of DreamHuman | text-to-3D avatar · realistic humans | CLIP score: 29.54 -> 32.17 | Improved |
| reduce training steps from 10,000 to 5,000 coarse and 3,000 to 2,000 fine (ablation) | AvatarStudio · optimization time | hours: 2.5 hours -> around 1 hour | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- AvatarVerse — AvatarVerse: High-Quality & Stable 3D Avatar Creation from Text and Pose (AAAI 2024) · better on 1 of 1 shared result
- DreamWaltz — DreamWaltz: Make a Scene with Complex 3D Animatable Avatars (NeurIPS 2023) · better on 1 of 1 shared result
- DreamHuman — DreamHuman: Animatable 3D Avatars from Text (NeurIPS 2023) · better on 1 of 1 shared result
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
- AvatarVerse: High-Quality & Stable 3D Avatar Creation from Text and Pose (AAAI 2024 · similarity 0.92)
- X-Oscar: A Progressive Framework for High-quality Text-guided 3D Animatable Avatar Generation (ICML 2024 · similarity 0.90)
- GAvatar: Animatable 3D Gaussian Avatars with Implicit Mesh Learning (CVPR 2024 · similarity 0.89)
- Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation (CVPR 2024 · similarity 0.89)
- DreamAvatar: Text-and-Shape Guided 3D Human Avatar Generation via Diffusion Models (CVPR 2024 · similarity 0.89)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.89)
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.
