Research map / 3D vision / 3D generation
Autodecoding Latent 3D Diffusion Models
Evangelos Ntavelis, Aliaksandr Siarohin, Kyle Olszewski, Chaoyang Wang, Luc Van Gool, Sergey Tulyakov
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 autodecoder latent diffusion instead of DiffRF | 3D generation · PhotoShape Chairs | FID: 15.95 -> 11.28 | Improved |
| Use autodecoder latent diffusion instead of DiffRF | 3D generation · ABO Tables | FID: 27.06 -> 18.44 | Improved |
| Replace direct latent sampling with diffusion sampling | Autodecoding Latent 3D Diffusion · CelebV-Text | FID: 69.21 -> 48.01 | Improved |
| Replace direct latent sampling with diffusion sampling | Autodecoding Latent 3D Diffusion · MVImgNet | FID: 97.51 -> 43.85 | Improved |
| Replace direct latent sampling with diffusion sampling | Autodecoding Latent 3D Diffusion · Objaverse | FID: 72.76 -> 40.49 | Improved |
| Add multi-frame training (ablation) | Autodecoder · PhotoShape Chairs reconstruction | PSNR: 27.176 -> 27.719 | Improved |
| Add multi-frame training (ablation) | Autodecoder · PhotoShape Chairs reconstruction | LPIPS: 6.855 -> 6.255 | Improved |
| Add self-attention (ablation) | Autodecoder · PhotoShape Chairs reconstruction | PSNR: 27.335 -> 27.719 | Hurt |
| Add self-attention (ablation) | Autodecoder · PhotoShape Chairs reconstruction | LPIPS: 6.738 -> 6.255 | Hurt |
| Increase residual-block depth (ablation) | Autodecoder · PhotoShape Chairs reconstruction | PSNR: 27.24 -> 27.719 | Improved |
| Increase residual-block depth (ablation) | Autodecoder · PhotoShape Chairs reconstruction | LPIPS: 6.924 -> 6.255 | Improved |
| Increase embedding length from 64 to 1024 (ablation) | Autodecoder · PhotoShape Chairs reconstruction | PSNR: 25.985 -> 27.719 | Improved |
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.
- DiffRF — DiffRF: Rendering-Guided 3D Radiance Field Diffusion (CVPR 2023) · better on 4 of 4 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
- LT3SD: Latent Trees for 3D Scene Diffusion (CVPR 2025 · similarity 0.90)
- LaGeM: A Large Geometry Model for 3D Representation Learning and Diffusion (ICLR 2025 · similarity 0.90)
- WildFusion: Learning 3D-Aware Latent Diffusion Models in View Space (ICLR 2024 · similarity 0.89)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.89)
- Generating Images with 3D Annotations Using Diffusion Models (ICLR 2024 · similarity 0.88)
- Denoising Diffusion via Image-Based Rendering (ICLR 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.
