Research map / 3D vision / Face and shape
One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization
Minghua Liu, Chao Xu, Haian Jin, Linghao Chen, Mukund Varma T, Zexiang Xu, Hao Su
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 Zero123 multi-view predictions with feed-forward SDF reconstruction instead of per-shape optimization | One-2-3-45 · Objaverse and GSO | F-score | Improved |
| Use view-conditioned Zero123 predictions to preserve the input image | One-2-3-45 · Single-image 3D reconstruction | input-image adherence | Improved |
| Use view-conditioned multi-view predictions rather than independently optimizing each view | One-2-3-45 · Single-image 3D reconstruction | 3D consistency | Improved |
| Use SDF representation instead of density-field reconstruction | One-2-3-45 · Single-image 3D reconstruction | geometry quality | Improved |
| Use 2-stage source view selection instead of uniformly posed Zero123 source views (ablation) | One-2-3-45 reconstruction module · 360-degree reconstruction | reconstruction quality | Improved |
| Use four nearby views in addition to the eight broader source views (ablation) | One-2-3-45 reconstruction module · 360-degree reconstruction | fine-grained geometry | Improved |
| Add depth loss during training (ablation) | One-2-3-45 reconstruction module · 360-degree reconstruction | fine-grained geometry | Improved |
| Train with ground-truth views plus nearby Zero123 predictions rather than only ground-truth views (ablation) | One-2-3-45 reconstruction module · Zero123-prediction inference | reconstruction completeness | Improved |
| Keep ground-truth views for depth supervision rather than replacing them with Zero123 predictions (ablation) | One-2-3-45 reconstruction module · 360-degree reconstruction | reconstruction quality | Improved |
| Estimate the input-view elevation rather than offsetting source-view elevations by ±30° (ablation) | One-2-3-45 · 360-degree reconstruction | shape distortion | Improved |
| Vary the number of source views and retrain the reconstruction module for that setting (ablation) | One-2-3-45 reconstruction module · 360-degree reconstruction | reconstruction quality | No clear effect |
Later papers that compared with this method
12 later accepted papers report One-2-3-45++ as a baseline in their results tables; 12 of them report a better number on most of the results they share with it.
- Multimodal Semantic Bias Mitigation for Diverse Text-To-3D Generation (CVPR 2026 · better on 3 of 3 shared results)
- ArchiSet: Benchmarking Editable and Consistent Single-View 3D Reconstruction of Buildings with Specific Window-to-Wall Ratios (ICCV 2025 · better on 4 of 6 shared results)
- High-fidelity 3D Object Generation from Single Image with RGBN-Volume Gaussian Reconstruction Model (CVPR 2025 · better on 4 of 4 shared results)
- Gamba: Marry Gaussian Splatting With Mamba for Single-View 3D Reconstruction (TPAMI 2025 · better on 2 of 2 shared results)
- Wonder3D++: Cross-Domain Diffusion for High-Fidelity 3D Generation From a Single Image (TPAMI 2025 · better on 1 of 1 shared result)
- MicroDreamer: Efficient 3D Generation in $\sim$20 Seconds by Score-based Iterative Reconstruction (TPAMI 2025 · better on 1 of 1 shared result)
- Wonder3D: Single Image to 3D using Cross-Domain Diffusion (CVPR 2024 · better on 2 of 2 shared results)
- SyncDreamer: Generating Multiview-consistent Images from a Single-view Image (ICLR 2024 · better on 2 of 2 shared results)
4 more not listed.
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
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · similarity 0.93)
- Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors (ICLR 2024 · similarity 0.90)
- Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image (NeurIPS 2024 · similarity 0.89)
- Wonder3D: Single Image to 3D using Cross-Domain Diffusion (CVPR 2024 · similarity 0.89)
- DIFIX3D+: Improving 3D Reconstructions with Single-Step Diffusion Models (CVPR 2025 · similarity 0.89)
- MeshFormer : High-Quality Mesh Generation with 3D-Guided Reconstruction Model (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.
