HawkFranklin-Research/SCIN-Dermatology-Raw-Images

SCIN-Dermatology-Raw-Images

This dataset contains 6,517 patient-submitted photographs organized into 3,061 clinical cases of common skin diseases. The source images are curated from the public Google Skin Condition Image Network (SCIN) corpus, cleansed of quality and gradability conflicts, and paired with complete patient-reported demographics, clinical symptoms, and dermatologist gradings.

Dataset Structure

This repository follows the standard Hugging Face ImageFolder structure:

.
├── README.md
├── metadata.csv
└── images/
    ├── -3205742176803893704.png
    ├── -4762289084741430925.png
    └── ...

The metadata.csv matches the image filenames in images/ to the respective case-level metadata and diagnostic splits.

Metadata Schema

Quick Start (Python)

To load and stream the images and metadata directly using the Hugging Face datasets library:

from datasets import load_dataset

# Load the dataset (automatically casts the 'image' field to PIL images)
dataset = load_dataset("HawkFranklin-Research/SCIN-Dermatology-Raw-Images")

# Access the training split
train_data = dataset["train"]

# Get the first sample
sample = train_data[0]
print(sample["primary_diagnosis"])  # e.g., 'Eczema'
print(sample["fitzpatrick_skin_type"])  # e.g., 'FST4'
sample["image"].show()  # Opens the PIL image

Citation & License

This dataset is distributed under the MIT License. The underlying image data and medical attributes are derived from the Google SCIN repository. If you use this dataset in your research, please cite the primary SCIN work:

@article{Ward2024_SCIN_JAMANetworkOpen,
  author    = {Ward, Edwin and others},
  title     = {Skin Condition Image Network (SCIN): A diverse dataset of patient-submitted skin photographs},
  journal   = {JAMA Network Open},
  year      = {2024},
  volume    = {7},
  number    = {5},
  pages     = {e2410389}
}