Research map / 3D vision / 3D generation
MicroDreamer: Efficient 3D Generation in $\sim$20 Seconds by Score-based Iterative Reconstruction
Luxi Chen, Zhengyi Wang, Zihan Zhou, Tingting Gao, Hang Su, Jun Zhu, Chongxuan Li
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 SIR instead of SDS | NeRF generation · NeRF with MVDream, Stable Zero123, ImageDream | speedup: 5-20 times | Improved |
| use MicroDreamer-20 iter instead of DreamGaussian | 3DGS mesh · Image-to-3D, 87 images | CLIP similarity: 0.692 -> 0.697 | Improved |
| use MicroDreamer-20 iter instead of DreamGaussian | 3DGS mesh · Image-to-3D, 87 images | generation time: ~30+10s -> ~18+3s | Improved |
| use MicroDreamer-30 iter instead of DreamGaussian | 3DGS mesh · Image-to-3D, 87 images | CLIP similarity: 0.692 -> 0.711 | Improved |
| use MicroDreamer-20 iter instead of DreamGaussian-300 iter at similar time | 3DGS mesh · Image-to-3D, 87 images | CLIP similarity: 0.641 -> 0.697 | Improved |
| increase MicroDreamer iterations from 20 to 30 | MicroDreamer · Image-to-3D, 87 images | CLIP similarity: 0.697 -> 0.711 | Improved |
| optimize SIR in pixel space instead of latent space (ablation) | SIR on NeRF · NeRF generation | speedup: 2-3 times | Improved |
| apply SDS in pixel space (ablation) | SDS · 3D generation | quality | Hurt |
| use iterative reconstruction (K>0) instead of K=0 initialization only (ablation) | SIR · 3DGS | generation quality | Improved |
| increase K beyond 30 (ablation) | SIR · 3DGS | generation quality | No clear effect |
| use more camera poses than 1 (ablation) | SIR · 3DGS with Stable Zero123 | generation quality | Improved |
| use linear time schedule instead of random or square schedule (ablation) | SIR · 3DGS | generation quality | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DreamGaussian — DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation (ICLR 2024) · better on 1 of 1 shared result
- One-2-3-45 — One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization (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
- IM-3D: Iterative Multiview Diffusion and Reconstruction for High-Quality 3D Generation (ICML 2024 · similarity 0.88)
- Hash3D: Training-free Acceleration for 3D Generation (CVPR 2025 · similarity 0.88)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.88)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · similarity 0.87)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.87)
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · 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.
