Dataset Viewer
The dataset viewer is not available for this subset.
The conch_v15_10x config contains 32 while it should generally contain 3 splits maximum (train/validation/test). If the splits TCGA_ACC, TCGA_BLCA, TCGA_BRCA, TCGA_CESC, TCGA_CHOL... are not used to differentiate between training and evaluation, please consider defining configs of this dataset instead. You can find how to define configs instead of splits here: https://huggingface.co/docs/hub/datasets-data-files-configuration
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
TCGA Pathology Feature Atlas
This repository stores pathology foundation-model embeddings as sharded Parquet files.
Physical layout:
data/{project_id}/{model_name}/{magnification}/part-*.parquet
Example:
data/TCGA-BRCA/conch_v15/20x/part-00000.parquet
Example streaming load:
from datasets import load_dataset
ds = load_dataset("IUCompPath/CRAFT", name="conch_v15_20x", split="TCGA_BRCA", streaming=True)
Each row is a chunk of patch embeddings from one slide.
- Downloads last month
- 58