Research map / Image generation / Editing and style transfer
RODIN: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion
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 RODIN instead of EG3D | 3D avatar generation · Unconditional generation | FID: 40.5 -> 26.1 | Helped |
| add latent conditioning to the plain 2D CNN baseline (ablation) | RODIN base diffusion model · Unconditional generation | FID: 39.2 -> 37.4 | Helped |
| add tri-plane roll-out to the latent-conditioned baseline (ablation) | RODIN base diffusion model · Unconditional generation | FID: 37.4 -> 28.4 | Helped |
| add 3D-aware convolution to rolled-out tri-planes (ablation) | RODIN base diffusion model · Unconditional generation | FID: 28.4 -> 26.1 | Helped |
| add image-level loss to tri-plane-level loss (ablation) | RODIN upsampler · Tri-plane upsampling | FID: 48.5 -> 38.6 | Helped |
| add condition augmentation to tri-plane and image-level losses (ablation) | RODIN upsampler · Tri-plane upsampling | FID: 38.6 -> 26.1 | Helped |
| use CFG scale 1.2 instead of disabling CFG (ablation) | RODIN conditional reconstruction · Conditional avatar reconstruction | LPIPS: 0.128 -> 0.121 | Helped |
| increase CFG scale from 1.2 to 6.0 (ablation) | RODIN conditional reconstruction · Conditional avatar reconstruction | SSIM: 0.794 -> 0.775 | Hurt |
Closest papers by meaning
- DiffusionGAN3D: Boosting Text-guided 3D Generation and Domain Adaptation by Combining 3D GANs and Diffusion Priors (CVPR 2024 · similarity 0.92)
- Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation (CVPR 2024 · similarity 0.91)
- AvatarArtist: Open-Domain 4D Avatarization (CVPR 2025 · similarity 0.89)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.89)
- DiffusionAvatars: Deferred Diffusion for High-fidelity 3D Head Avatars (CVPR 2024 · similarity 0.89)
- DIRECT-3D: Learning Direct Text-to-3D Generation on Massive Noisy 3D Data (CVPR 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.