Hospital Model Hub
Audited foundation models and public medical datasets, available on the hospital intranet
Models · Departments
Models are assigned by the data types each department commonly uses (one rule for all models; general-purpose models are listed separately). Datasets are assigned by where and on what anatomy the data were acquired.
Gastroenterology / Endoscopy 4ENT / Head & Neck 6Pathology 38Radiology 20Oncology 51Surgery 4Ophthalmology 7Dermatology 5Cardiology 7Neurology 9Obstetrics / Ultrasound 3Critical care / Emergency 18Laboratory / Genetics 5General (all departments) 101
Datasets · Departments
Gastroenterology / Endoscopy 11ENT / Head & Neck 6Pathology 16Radiology 38Oncology 32Surgery 7Ophthalmology 11Dermatology 5Cardiology 6Neurology 7Obstetrics / Ultrasound 5Critical care / Emergency 6Laboratory / Genetics 2General (all departments) 1
Models · Domains
General 86Pathology 38Medical (general) 10Chest X-ray 83D CT 7Ophthalmology 7Time series 5ECG 5Molecular / Omics 5Video 5CT / MRI 5Dermatology 5EEG 4Surgical video 3Echocardiography 2Endoscopy 1Fetal ultrasound 1
Categories
Most downloaded
openai/clip-vit-large-patch14
Disclaimer: The model card is taken and modified from the official CLIP repository, it can be found here.
facebook/esm2_t33_650M_UR50D
ESM-2 is a state-of-the-art protein model trained on a masked language modelling objective. It is suitable for fine-tuning on a wide range o…
facebook/dinov2-base
Vision Transformer (ViT) model trained using the DINOv2 method. It was introduced in the paper DINOv2: Learning Robust Visual Features witho…
facebook/dinov2-small
Vision Transformer (ViT) model trained using the DINOv2 method. It was introduced in the paper DINOv2: Learning Robust Visual Features witho…
IDEA-Research/grounding-dino-base
The Grounding DINO model was proposed in Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection by Shilong L…
google/siglip2-giant-opt-patch16-384
SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved sema…
facebook/sam3
SAM 3 is a unified foundation model for promptable segmentation in images and videos. It can detect, segment, and track objects using text o…
google/siglip2-base-patch16-224
SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved sema…
How to use
- Browse models and datasets by department, domain, category (task) and access; sort by downloads or size.
- Search: Chinese medical terms are expanded to English; any keyword may match; ranked over full card text.
- Details: model and dataset pages show the authors' original cards; dataset pages also state provenance and the original release.
- Request: call hub_request(entry_id, reason). Staff download manually; after requesting, pause and report.
- Agents open any page with hub_browse(url); links in a page can be passed as url.