Research map / 3D vision / 3D generation
GeoWizard: Unleashing the Diffusion Priors for 3D Geometry Estimation from a Single Image
Xiao Fu, Wei Yin, Mu Hu, Kaixuan Wang, Yuexin Ma, Ping Tan, Shaojie Shen, Dahua Lin, Xiaoxiao Long
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 |
|---|---|---|---|
| add cross-domain geometry switcher and cross-domain attention for joint depth and normal (ablation) | GeoWizard · Ablation validation sets (overall) | Geometric consistency (GC): 18.1 -> 16.2 | Improved |
| use one joint model instead of separate depth and normal models (ablation) | GeoWizard · Ablation validation sets (overall) | AbsRel: 8.5 -> 6.7 | Improved |
| add scene distribution decoupler (ablation) | GeoWizard · Ablation validation sets (overall) | AbsRel: 7.5 -> 6.7 | Improved |
| add scene distribution decoupler (ablation) | GeoWizard · Object domain (OmniObject3D samples) | AbsRel: 3.7 -> 3.5 | No clear effect |
| use geometry switcher instead of shared geometry embedding (ablation) | GeoWizard · Ablation validation sets (overall) | AbsRel: 7.2 -> 6.7 | Improved |
| use wrong domain indicator at test time (indoor test with outdoor indicator) (ablation) | GeoWizard · Indoor ablation set | Geometric consistency (GC): 14.7 -> 14.4 | Mixed |
| use wrong domain indicator at test time (object indicator on indoor) (ablation) | GeoWizard · Indoor ablation set | AbsRel: 5.5 -> 6.4 | Hurt |
| use GeoWizard depth/normal as guidance for MonoSDF reconstruction instead of Omnidata v2 or DSINE | MonoSDF · ScanNet (4 scenes) | F-score: 74.7 -> 75.1 | Improved |
| generative diffusion-prior model versus discriminative DepthAnything trained on 63.5M images | GeoWizard · NYUv2 (zero-shot depth) | AbsRel: 4.3 -> 5.2 | Hurt |
Later papers that compared with this method
5 later accepted papers report GeoWizard as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Monocular Normal Estimation via Shading Sequence Estimation (ICLR 2026 · better on 3 of 3 shared results)
- Neural LightRig: Unlocking Accurate Object Normal and Material Estimation with Multi-Light Diffusion (CVPR 2025 · better on 1 of 1 shared result)
- IDArb: Intrinsic Decomposition for Arbitrary Number of Input Views and Illuminations (ICLR 2025 · better on 1 of 1 shared result)
- Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation (ICCV 2025 · better on 1 of 1 shared result)
- GeoMan: Temporally Consistent Human Geometry Estimation using Image-to-Video Diffusion (ICCV 2025 · better on 1 of 1 shared result)
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.
- Omnidata v2 — Omnidata: A Scalable Pipeline for Making Multi-Task Mid-Level Vision Datasets From 3D Scans (ICCV 2021) · 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
- GeoComplete: Geometry-Aware Diffusion for Reference-Driven Image Completion (NeurIPS 2025 · similarity 0.88)
- GeoVideo: Introducing Geometric Regularization into Video Generation Model (NeurIPS 2025 · similarity 0.87)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.87)
- GeoDepth: From Point-to-Depth to Plane-to-Depth Modeling for Self-Supervised Monocular Depth Estimation (CVPR 2025 · similarity 0.87)
- LaGeM: A Large Geometry Model for 3D Representation Learning and Diffusion (ICLR 2025 · similarity 0.87)
- Geometry Image Diffusion: Fast and Data-Efficient Text-to-3D with Image-Based Surface Representation (ICLR 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.
