Multi-Modal Foundation · MahmoodLab
UNI
A self-supervised Vision Transformer patch encoder for computational pathology and histology workflows.
What UNI is
UNI is a pathology foundation model developed by MahmoodLab for applications across medicine and biology. It provides general-purpose feature representations for computational pathology workflows by encoding histology image patches.
Histopathological image analysis requires robust representation of complex tissue morphology. UNI provides a pretrained visual foundation derived from large-scale tissue microscopy to support downstream computational pathology tasks.
How it works
UNI is based on a Vision Transformer (ViT) architecture trained using self-supervised learning.
Key facts
- Architecture
- Vision Transformer (ViT)
- Learning paradigm
- Self-supervised learning
- Primary task
- Image feature extraction
- Framework
- PyTorch (timm)
What you can use it for
- Patch Feature Extraction
- Encodes histology image patches into numerical feature representations for downstream machine learning workflows.
- H&E Whole-Slide Image Analysis
- Serves as an encoder backbone for downstream analytical pipelines operating on haematoxylin and eosin stained whole-slide images.
- Computational Pathology Research
- Provides a foundation model checkpoint for developing and evaluating biological and medical image analysis algorithms.
Run it here
We list this model but cannot run it yet: excluded (gated): Gated and CC-BY-NC-ND-4.0..
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Functions as a general-purpose patch encoder across computational pathology tasks.
- Utilises self-supervised representations without requiring manual patch-level annotations.
Watch out for
- Functions primarily as a patch-level feature extractor rather than an end-to-end slide-level classifier.
- Trained specifically on H&E whole-slide images and may not generalise directly to alternative staining techniques.
Inputs and outputs
You provide
Image
PNG, JPEG or TIFF. Sent to the model as-is.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · cc-by-nc-nd-4.0
This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
Questions
Can I use UNI commercially?
Check the licence before commercial use. This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
What is the primary function of UNI?
UNI functions as a general-purpose patch encoder that extracts feature representations from histology images for computational pathology workflows.
What architecture and training approach does UNI use?
UNI uses a Vision Transformer (ViT) architecture trained with self-supervised learning.
Written from UNI's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.