{"id":28505095,"url":"https://github.com/prs-eth/marigold-dc","last_synced_at":"2026-03-11T10:06:30.116Z","repository":{"id":268679095,"uuid":"900123818","full_name":"prs-eth/Marigold-DC","owner":"prs-eth","description":"[ICCV 2025] Zero-Shot Monocular Depth Completion with Guided Diffusion","archived":false,"fork":false,"pushed_at":"2025-10-16T15:07:38.000Z","size":3844,"stargazers_count":201,"open_issues_count":0,"forks_count":12,"subscribers_count":4,"default_branch":"main","last_synced_at":"2025-10-16T21:29:10.030Z","etag":null,"topics":["depth-completion","depth-estimation","depth-fusion","diffusion"],"latest_commit_sha":null,"homepage":"https://marigolddepthcompletion.github.io/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/prs-eth.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.txt","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2024-12-07T23:24:38.000Z","updated_at":"2025-10-16T15:07:37.000Z","dependencies_parsed_at":"2024-12-18T09:31:26.134Z","dependency_job_id":"e5182d99-d3b4-4af4-888e-35e9bbec2d57","html_url":"https://github.com/prs-eth/Marigold-DC","commit_stats":null,"previous_names":["prs-eth/marigold-dc"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/prs-eth/Marigold-DC","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prs-eth%2FMarigold-DC","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prs-eth%2FMarigold-DC/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prs-eth%2FMarigold-DC/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prs-eth%2FMarigold-DC/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/prs-eth","download_url":"https://codeload.github.com/prs-eth/Marigold-DC/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prs-eth%2FMarigold-DC/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":30377838,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-03-11T06:09:32.197Z","status":"ssl_error","status_checked_at":"2026-03-11T06:09:17.086Z","response_time":84,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["depth-completion","depth-estimation","depth-fusion","diffusion"],"created_at":"2025-06-08T19:06:38.210Z","updated_at":"2026-03-11T10:06:28.231Z","avatar_url":"https://github.com/prs-eth.png","language":"Python","readme":"# ⇆ Marigold-DC: Zero-Shot Monocular Depth Completion with Guided Diffusion (ICCV 2025)\n\n[![Website](https://img.shields.io/badge/%F0%9F%A4%8D%20Project%20-Website-blue)](https://marigolddepthcompletion.github.io)\n[![arXiv](https://img.shields.io/badge/arXiv-PDF-b31b1b)](http://arxiv.org/abs/2412.13389)\n[![Hugging Face Space](https://img.shields.io/badge/🤗%20Hugging%20Face%20-Space-yellow)](https://huggingface.co/spaces/prs-eth/marigold-dc)\n[![License](https://img.shields.io/badge/License-Apache--2.0-929292)](https://www.apache.org/licenses/LICENSE-2.0)\n\nThis repository represents the official implementation of the paper titled \"Marigold-DC: Zero-Shot Monocular Depth Completion with Guided Diffusion\".\n\nPhotogrammetry and Remote Sensing team: \n[Massimiliano Viola](https://www.linkedin.com/in/massimiliano-viola/), \n[Kevin Qu](https://www.linkedin.com/in/kevin-qu-b3417621b/), \n[Nando Metzger](https://nandometzger.github.io/), \n[Bingxin Ke](http://www.kebingxin.com/),\n[Alexander Becker](https://scholar.google.ch/citations?user=Wle2GmkAAAAJ\u0026hl=en), \n[Konrad Schindler](https://scholar.google.com/citations?user=FZuNgqIAAAAJ\u0026hl=en),\n[Anton Obukhov](https://www.obukhov.ai/).\n\n![](doc/teaser.jpg)\n\n## 🛠️ Setup\n\n📦 Clone the repository:\n```bash\ngit clone https://github.com/prs-eth/Marigold-DC.git\ncd Marigold-DC\n```\n\n🐍 Create python environment:\n```bash\npython -m venv venv/marigold_dc\n```\n\n⚡ Activate the environment:\n```bash\nsource venv/marigold_dc/bin/activate\n```\n\n💻 Install the dependencies:\n```bash\npip install -r requirements.txt\n```\n\n## 🚀 Usage\n\nThe script performs densification of the input sparse depth, provided as a sparse numpy array, \nand saves the output as a dense numpy array, along with the visualization. \nOptimal default settings are applied.\nBy default, it processes the [teaser image](data/image.png) and uses [100-point guidance](data/sparse_100.npy). \n\n🏃🏻‍♂️‍➡️ Simply run as follows:\n```bash\npython -m marigold_dc\n```\n\n🧩 Customize image and sparse depth inputs as follows:\n```bash\npython -m marigold_dc \\\n    --in-image \u003cPATH_RGB_IMAGE\u003e \\\n    --in-depth \u003cPATH_SPARSE_DEPTH\u003e \\\n    --out-depth \u003cPATH_DENSE_DEPTH\u003e\n```\n\n🛠️ Customize other settings:\n- `--num_inference_steps \u003cint\u003e` specifies the number of diffusion inference steps.\n- `--checkpoint \u003cpath\u003e` allows overriding the base monocular depth estimation model checkpoint; can be a local path or a Hugging Face repository.\n\n## 🏋️‍♂️ Training\n\nNone — the method is purely test-time; please refer to the paper for more details.\n\n## ⬇ Checkpoint cache\nBy default, the [checkpoint](https://huggingface.co/prs-eth/marigold-depth-v1-0) is stored in the Hugging Face cache, \nwhich defaults to the home directory on Linux and Mac. \nThis is often problematic in cluster environments.\nThe `HF_HOME` environment variable defines the cache location and can be overridden, e.g.:\n\n```\nexport HF_HOME=/large_volume/cache\n```\n\n## 🦿 Evaluation on test datasets\nComing soon\n\n## Abstract\n\nDepth completion upgrades sparse depth measurements into dense depth maps, guided by a conventional image. \nExisting methods for this highly ill-posed task operate in tightly constrained settings, \nand tend to struggle when applied to images outside the training domain, \nas well as when the available depth measurements are sparse, irregularly distributed, or of varying density. \nInspired by recent advances in monocular depth estimation, \nwe reframe depth completion as image-conditional depth map generation, guided by a sparse set of measurements. \nOur method, Marigold-DC, builds on a pretrained latent diffusion model (LDM) for depth estimation and injects \nthe depth observations as test-time guidance, via an optimization scheme that runs in tandem with the iterative \ninference of denoising diffusion. The method exhibits excellent zero-shot generalization across a diverse range \nof environments and handles even extremely sparse guidance effectively. Our results suggest that contemporary \nmonodepth priors greatly robustify depth completion: it may be better to view the task as recovering dense depth \nfrom (dense) image pixels, guided by sparse depth; rather than as inpainting (sparse) depth, guided by an image.\n\n## 📢 News\n\n - 2024-12-19: ArXiv paper and demo release.\n - 2024-12-18: Code release (this repository).\n\n## 🎓 Citation\n```bibtex\n@misc{viola2024marigolddc,\n    title={Marigold-DC: Zero-Shot Monocular Depth Completion with Guided Diffusion}, \n    author={Massimiliano Viola and Kevin Qu and Nando Metzger and Bingxin Ke and Alexander Becker and Konrad Schindler and Anton Obukhov},\n    year={2024},\n    eprint={2412.13389},\n    archivePrefix={arXiv},\n    primaryClass={cs.CV},\n}\n```\n\n## 🎫 License\n\nThe code of this work is licensed under the Apache License, Version 2.0 (as defined in the [LICENSE](LICENSE.txt)).\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fprs-eth%2Fmarigold-dc","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fprs-eth%2Fmarigold-dc","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fprs-eth%2Fmarigold-dc/lists"}