vit_small_patch16_256.tcga_brca_dino

Model card for vit_small_patch16_256.tcga_brca_dino

A Vision Transformer (ViT) image classification model.
Trained on 2M histology patches from TCGA-BRCA.

Model Details

• Model Type: Feature backbone
• Model Stats:
  • Params (M): 21.7
  • Image size: 256 x 256 x 3
• Papers:
  • Self-Supervised Vision Transformers Learn Visual Concepts in Histopathology: https://arxiv.org/abs/2203.00585
• Dataset: TCGA BRCA: https://portal.gdc.cancer.gov/
• Original: https://github.com/Richarizardd/Self-Supervised-ViT-Path/
• License: GPLv3

Model Usage

Image Embeddings

from urllib.request import urlopen
from PIL import Image
import timm

# get example histology image
img = Image.open(
  urlopen(
    "https://github.com/owkin/HistoSSLscaling/raw/main/assets/example.tif"
  )
)

# load model from the hub
model = timm.create_model(
  model_name="hf-hub:1aurent/vit_small_patch16_256.tcga_brca_dino",
  pretrained=True,
).eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # output is (batch_size, num_features) shaped tensor

Citation

@misc{chen2022selfsupervised,
  title         = {Self-Supervised Vision Transformers Learn Visual Concepts in Histopathology},
  author        = {Richard J. Chen and Rahul G. Krishnan},
  year          = {2022},
  eprint        = {2203.00585},
  archiveprefix = {arXiv},
  primaryclass  = {cs.CV}
}