Home / AI Models / MAD-NP
Model released image-feature-extraction apache-2.0

MAD-NP

🤗 Hugging Face

Details

Architecture ViT-Giant with Registers (~1B params)
Parameters ~1B
Base Model facebook/dinov2-with-registers-giant
Relation finetune
License apache-2.0

Magnification-Aware Distillation for Neuropathology. This model maintains consistent representations across magnification levels through a novel magnification-aware alignment loss.

Results

MetricValue
Linear F10.9307
KNN F10.9286
AMI0.7668

Outperforms UNI2, Prov-GigaPath, UNI, Virchow2.

Related

Publications

Publication Magnification-Aware Distillation (MAD): A Self-Supervised Framework for Unified Representation Learning in Gigapixel Whole-Slide Images

Grants

Grant Federated Digital Pathology Platform for AD/ADRD Research

Related Datasets

Dataset Neuropathology WSI Dataset

Platforms Used

Platform Vision Foundry