# MedMNIST/MedMNIST

[pip install medmnist] 18x Standardized Datasets for 2D and 3D Biomedical Image Classification

Repository: https://github.com/MedMNIST/MedMNIST
Canonical: https://ross.abutalabs.com/products/medmnist
Homepage: https://medmnist.com/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: dataset, benchmark, automl, mnist, medical, medical-image-analysis, medmnist, multi-modal, decathlon, medical-imaging, medical-image-computing, deep-learning, machine-learning, 3d, 2d, classification, few-shot-learning, pytorch, federated-learning
Last push: 2025-01-11T14:34:49+00:00

## Health v2 (maintenance only)
Score: 33/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 1, release rhythm 35, longevity 100
- inputs: {"age_days": 2138, "days_push": 599, "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 1398, forks 213 (observed 2026-08-28T04:04:36.808769+00:00)

## What it is
MedMNIST is a collection of 18 standardized biomedical image classification datasets (12 2D and 6 3D), available in multiple sizes from MNIST-like 28px up to 224px, with a pip-installable Python API. It serves as a lightweight benchmark for medical image analysis, machine learning research, and education.

## Use cases
- benchmark medical image classification models
- find a lightweight MNIST-like dataset for biomedical deep learning
- get standardized 2D and 3D medical images for teaching machine learning
- evaluate AutoML tools on medical imaging tasks
- train models on multi-label and ordinal regression medical image tasks
- benchmark medical foundation models on larger 64/128/224 images

## When to choose
- you need standardized, ready-to-use biomedical image datasets without domain preprocessing
- you want a lightweight benchmark covering diverse modalities, scales, and task types
- you are doing education, prototyping, or fair model comparison in medical imaging

## When to avoid
- you need full-resolution clinical images or raw DICOM data
- you need pixel-level segmentation or detection annotations rather than classification labels
- your production system requires real-world clinical validation data

## Facets
- artifact type: dataset
- maturity: stable
- function: machine-learning, deep-learning, image-processing, benchmarking, data-science
- domain: machine-learning, computer-vision, healthcare, bioinformatics, artificial-intelligence
- platform: python, cross-platform
- tags: medical-imaging, mnist-like, image-classification, biomedical, pytorch, automl-benchmark, 2d-3d-images, few-shot-learning

## Member repositories
- MedMNIST/MedMNIST (main) score 33

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:36.808769+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 2026-08-30T04:39:12.301952+00:00, confidence not recorded.
  - readme: https://github.com/MedMNIST/MedMNIST (fetched 2026-08-28T04:04:36.808769+00:00, sha ec86f062dbb4)
  - homepage: https://medmnist.com/ (fetched 2026-08-29T11:53:29.002006+00:00, sha 147e2b38747a)
  - registry_pypi: https://pypi.org/pypi/medmnist/json (fetched 2026-08-29T11:53:29.005090+00:00, sha 2912075934b0)
- Data as of 2026-08-30T08:39:29.467469+00:00.