{"data":{"slug":"mrgiovanni-modelsgenesis","name":"ModelsGenesis","tagline":"Foundation models for medical image analysis","github_url":"https://github.com/MrGiovanni/ModelsGenesis","owner":"MrGiovanni","repo":"ModelsGenesis","owner_avatar_url":"https://avatars.githubusercontent.com/u/9360531?v=4","primary_language":"Jupyter Notebook","stars":791,"forks":141,"topics":["3d-model","3d-segmentation","chest-ct","ct","feature-learning","fine-tuning","foundation-models","medical-image-analysis","medical-imaging","mr","pre-trained-model","representation-learning","self-supervised-learning","transfer-learning","unsupervised-learning"],"archived":false,"github_pushed_at":"2026-08-25T17:16:38+00:00","maintenance_label":"Active","stars_delta_30d":3,"url":"https://www.graphcanon.com/tools/mrgiovanni-modelsgenesis","markdown_url":"https://www.graphcanon.com/tools/mrgiovanni-modelsgenesis.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/mrgiovanni-modelsgenesis","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=mrgiovanni-modelsgenesis","description":"[MICCAI 2019 Young Scientist Award] [MedIA Best Paper Award] Models Genesis: self-supervised pre-training for 3D medical images. Learns transferable representations from unlabeled CT and MRI volumes, then fine-tunes for downstream segmentation and classification. Keras and PyTorch weights included.","homepage_url":"https://www.cs.jhu.edu/~zongwei/publication/zhou2020models.pdf","license":"Other","open_issues":28,"watchers":13,"ai_summary":"ModelsGenesis is among the earliest foundation models in medical imaging, supporting multiple downstream tasks through pre-trained representations.","readme_excerpt":"

\n\n
\n\n\n\n\n
\n\n**Models Genesis** is a self-supervised learning framework that pre-trains 3D models on unlabeled medical images. It learns transferable visual representations from CT and MRI volumes with no manual annotation, and those representations fine-tune for downstream segmentation and classification.\n\nThis line of work is described in the literature as self-supervised learning, representation learning, feature learning, and transfer learning. Models Genesis is an early example of what would later be called a foundation model for medical imaging.\n\nPre-training uses image restoration proxy tasks: non-linear intensity transformation, local pixel shuffling, out-painting, and in-painting. The model learns anatomy by reconstructing deliberately distorted volumes. In the self-supervised learning literature this places Models Genesis in the *generative* or *restorative* family, as distinct from contrastive methods.\n\nThe original name, Generic Autodidactic Models, describes the same idea: models created ex nihilo (with no manual labeling), self-taught (learned by self-supervision), and generic (serving as source models for application-specific target models). Models Genesis is a primary source of transfer learning for 3D medical imaging, particularly where annotated data is limited. \n\n

\n

\n\n\n## Paper\nThis repository provides the official implementation of training Models Genesis as well as the usage of the pre-trained Models Genesis in the following paper:\n\nModels Genesis: Generic Autodidactic Models for 3D Medical Image Analysis
\n[Zongwei Zhou](https://www.zongweiz.com/)1, [Vatsal Sodha](https://github.com/vatsal-sodha)1, [Md Mahfuzur Rahman Siddiquee](https://github.com/mahfuzmohammad)1,
\n[Ruibin Feng](https://chs.asu.edu/ruibin-feng)1, [Nima Tajbakhsh](https://www.linkedin.com/in/nima-tajbakhsh-b5454376/)1, [Michael B. Gotway](https://www.mayoclinic.org/biographies/gotway-michael-b-m-d/bio-20055566)2, and [Jianming Liang](https://chs.asu.edu/jianming-liang)1
\n1 Arizona State University, 2 Mayo Clinic
\nInternational Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2019
\n[Young Scientist Award](https://miccai.org/conference-awards/best-paper-award-and-young-scientist-award/)
\n[paper](https://www.cs.toronto.edu/~liang/Publications/ModelsGenesis/MICCAI_2019_Full.pdf) | [code](https://github.com/MrGiovanni/ModelsGenesis) | [slides](https://docs.wixstatic.com/ugd/deaea1_c5e0f8cd9cde4c3db339d866483cbcd3.pdf) | [poster](https://www.cs.toronto.edu/~liang/Publications/ModelsGenesis/Models_Genesis_Poster.pdf) | talk ([YouTube](https://youtu.be/5W_uGzBloZs), [YouKu](https://v.youku.com/v_show/id_XNDM5NjQ1ODAxMg==.html?sharefrom=iphone&sharekey=496e1494c76ed263653aa3aada61c23e6)) | [blog](https://zhuanlan.zhihu.com/p/86366534)\n\nModels Genesis
\n[Zongwei Zhou](https://www.zongweiz.com/)1, [Vatsal Sodha](https://github.com/vatsal-sodha)1, [Jiaxuan Pang](https://github.com/MRJasonP)1, [Michael B. Gotway](https://www.mayoclinic.org/biographies/gotway-michael-b-m-d/bio-20055566)2, and [Jianming Liang](https://chs.asu.edu/jianming-liang)1
\n1 Arizona State University, 2 Mayo Clinic
\nMedical Image Analysis (MedIA)
\n[MedIA Best Paper Award](https://miccai.org/conference-awards/medical-image-analysis-best-paper-award/)
\n[paper](https://arxiv.org/pdf/2004.07882.pdf) | [code](https://github.com/MrGiovanni/ModelsGenesis) | [slides](https://d5b3ebbb-7f8d-4011-9114-d87f4a930447.filesusr.com/ugd/deaea1_5ecdfa48836941d6ad174d","github_created_at":"2019-07-24T19:07:19+00:00","created_at":"2026-07-11T11:40:25.8361+00:00","updated_at":"2026-09-20T06:30:43.688156+00:00","categories":[{"slug":"evaluation-observability","name":"Evaluation & Observability","url":"https://www.graphcanon.com/categories/evaluation-observability","markdown_url":"https://www.graphcanon.com/categories/evaluation-observability.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/evaluation-observability"},{"slug":"model-training","name":"Model Training","url":"https://www.graphcanon.com/categories/model-training","markdown_url":"https://www.graphcanon.com/categories/model-training.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/model-training"}],"tags":[{"slug":"3d-model","name":"3d-model"},{"slug":"fine-tuning","name":"fine-tuning"},{"slug":"foundation-models","name":"foundation models"},{"slug":"pre-trained-model","name":"pre-trained model"},{"slug":"representation-learning","name":"representation-learning"},{"slug":"self-supervised-learning","name":"self-supervised-learning"},{"slug":"transfer-learning","name":"transfer-learning"}],"trust":{"provenance":{"is_fork":false,"github_id":198694912,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-09-19T18:26:28.447Z","maintenance":{"label":"Active","score":82,"methodology":"github_public_v1","releases_90d":0,"days_since_push":25,"last_release_at":null,"stars_delta_30d":3,"open_issues_delta_30d":0},"security_summary":{"status":"no_lockfile","scanner":null,"low_count":0,"high_count":0,"last_scan_at":"2026-07-11T11:40:27.219Z","medium_count":0,"scan_profile":"none","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-09-19T18:26:29.832Z"},"languages":{"value":["jupyter notebook"],"source":"github.language","observed_at":"2026-09-19T18:26:29.832Z"},"license_spdx":{"value":"Other","source":"github.license","observed_at":"2026-09-19T18:26:29.832Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":{"notes":["Jupyter Notebook environment is required for leveraging ModelsGenesis pre-trained models and conducting fine-tuning."],"requires_docker":false},"constraints":{"requires_docker":false},"when_to_use":["If you are working on downstream tasks in medical image analysis where a robust foundation model enhances accuracy and reduces training time","When focusing on tasks requiring complex representation learning such as 3D models, ModelsGenesis provides a solid base for fine-tuning"],"when_not_to_use":["Avoid if your project requires real-time inference capabilities, as ModelsGenesis focuses more on improving model quality through extensive pre-training rather than optimizing for speed","Not recommended if you are in need of domain-general foundation models that work across various industries, as it is specialized strictly towards medical imaging"],"source":"enrich:decision_facts","observed_at":"2026-07-15T09:50:42.960Z"},"constraint_facets":{"requires_docker":false},"decision_summary":[{"label":"Requirements","value":"Jupyter Notebook environment is required for leveraging ModelsGenesis pre-trained models and conducting fine-tuning."},{"label":"Adopt for","value":"ModelsGenesis is notable for its foundational approach to pre-trained models specific to medical imaging tasks, with awards validating its contribution to the field of transfer learning and self-supervised strategies."},{"label":"License detail","value":"Other: The precise licensing terms must be verified directly from the tool's official documentation or repository to understand usage rights."}]}}