Research map / Image generation / Text-to-image diffusion
Instant 3D Human Avatar Generation using Image Diffusion Models
Nikos Kolotouros, Thiemo Alldieck, Enric Corona, Eduard Gabriel Bazavan, Cristian Sminchisescu
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 |
|---|---|---|---|
| feed dense GHUM body pose and shape rendering into the 3D reconstruction network (ablation) | AvatarPopUp 3D reconstruction · 100 prompts with random pose and shape | Chamfer distance to GHUM mesh: 8.6 -> 1.4 | Improved |
| fine-tune only encoder convolutional layers instead of the whole diffusion network (ablation) | Latent Diffusion text-to-image model · text prompts with unseen garments | diversity and text alignment (qualitative) | Improved |
| generate a back view with a diffusion model before 3D lifting | single-image 3D reconstruction · held-out scans | FID color back: 40.8 -> 20.0 vs retrained PHORHUM | Improved |
| use AvatarPopUp instead of retrained PHORHUM on 3D geometry metrics | single-image 3D reconstruction · held-out scans | Chamfer distance: 0.84 -> 0.91 | No clear effect |
| use feed-forward AvatarPopUp instead of optimization-based text-to-3D | DreamHuman and TADA · text-to-3D human generation | CLIP R-Precision and runtime: runtime 8h -> ~10s; R-Prec 0.68 -> 0.62 | Mixed |
| use 5 DDIM steps and 256^3 marching cubes instead of 50 steps and 512^3 (ablation) | AvatarPopUp · text-to-3D human generation | CLIP R-Precision (Color): 0.62 -> 0.58; runtime ~10s -> ~2s | Mixed |
Closest papers by meaning
- Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation (CVPR 2024 · similarity 0.92)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.91)
- DreamAvatar: Text-and-Shape Guided 3D Human Avatar Generation via Diffusion Models (CVPR 2024 · similarity 0.91)
- AvatarVerse: High-Quality & Stable 3D Avatar Creation from Text and Pose (AAAI 2024 · similarity 0.90)
- AvatarGO: Zero-shot 4D Human-Object Interaction Generation and Animation (ICLR 2025 · similarity 0.90)
- DiffusionGAN3D: Boosting Text-guided 3D Generation and Domain Adaptation by Combining 3D GANs and Diffusion Priors (CVPR 2024 · similarity 0.90)
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.
