microsoft/MoGe

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

2807 stars · 215 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MoGe is a deep learning model from Microsoft Research that recovers 3D geometry from a single open-domain image, predicting metric point maps, depth maps, normal maps, and camera field of view in one forward pass. It ships with inference code, pretrained weights, and a Hugging Face demo, achieving ~60ms latency per image on modern GPUs.

Use cases

• estimate depth maps from single photos
• recover 3D point clouds from one image
• predict surface normal maps from images
• reconstruct 3D geometry for single-view 3D reconstruction
• estimate camera field of view from an image
• build 3D assets from photos for graphics pipelines

When to choose

• you need fast, accurate monocular geometry or depth estimation from arbitrary images
• you want point maps, depth, and normals from a single unified model
• you need a research-grade model with pretrained weights and a demo

When to avoid

• you need multi-view or video-based 3D reconstruction with true global scale
• you cannot run GPU inference
• you need a lightweight CPU-only depth estimator

Facets

library · maturity active

machine-learning computer-vision image-processing deep-learning computer-vision machine-learning graphics artificial-intelligence python cross-platform monocular-depth-estimation 3d-reconstruction point-maps depth-maps normal-maps cvpr-2025 research-model inference gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="microsoft/MoGe")