# prs-eth/Marigold

[CVPR 2024 - Oral, Best Paper Award Candidate] Marigold: Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation

Repository: https://github.com/prs-eth/Marigold
Canonical: https://ross.abutalabs.com/products/marigold
Homepage: https://marigoldmonodepth.github.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: monocular-depth-estimation, diffusion, in-the-wild, zero-shot, depth-estimation
Last push: 2025-12-10T09:56:30+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 56, release rhythm 35, longevity 72
- inputs: {"age_days": 1010, "days_push": 266, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3198, forks 208 (observed 2026-08-28T04:07:48.367581+00:00)

## What it is
Marigold is a family of diffusion-based models and a fine-tuning protocol that adapts pretrained latent diffusion models like Stable Diffusion for dense image analysis tasks. It delivers state-of-the-art zero-shot monocular depth estimation, surface normal prediction, and intrinsic image decomposition.

## Use cases
- estimate depth maps from a single image
- predict surface normals from photos
- decompose an image into albedo and shading
- run zero-shot depth estimation on in-the-wild images
- fine-tune stable diffusion for dense prediction tasks
- generate high-resolution depth maps for 3d reconstruction

## When to choose
- you need high-quality monocular depth estimation with strong zero-shot generalization
- you want to repurpose pretrained diffusion models for dense image analysis
- you need surface normals or intrinsic decomposition from single images
- you want a model trainable on small synthetic datasets with a single GPU

## When to avoid
- you need real-time depth estimation on low-power or CPU-only hardware
- you need metric depth with calibrated scale rather than affine-invariant depth
- you want a lightweight classical depth estimator without diffusion inference overhead

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, image-processing, deep-learning, machine-learning
- platform: python, cross-platform
- tags: monocular-depth-estimation, diffusion-models, stable-diffusion, surface-normals, intrinsic-decomposition, zero-shot, cvpr-2024, huggingface, gpu

## Member repositories
- prs-eth/Marigold (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:48.367581+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:52.362151+00:00, confidence not recorded.
  - readme: https://github.com/prs-eth/Marigold (fetched 2026-08-28T04:07:48.367581+00:00, sha 92fef40e7c82)
  - homepage: https://marigoldmonodepth.github.io (fetched 2026-08-29T09:38:33.593204+00:00, sha 75368b8d20e2)
- Data as of 2026-08-30T08:39:29.467469+00:00.