z72pepee/dermatological-diagnoses

Dataset Card for dermatological diagnoses

This dataset is aimed at grouping and facilitating access to large volumes of data on the diagnosis of dermatological lesions. The datasets come from various sources and have been preprocessed and standardized so that they can be immediately used with your preferred model. The resolution of these is 224x224.

You can find 16 datasets that have been used in several articles and competitions on the subject, from 2 to 198 categories. The DERM-LIB dataset is not included as it requires a license that can be obtained here: https://licensing.edinburgh-innovations.ed.ac.uk/product/dermofit-image-library.

Below, we present examples of how you can download and use each of them.

How to use

from datasets import load_dataset

data_names = [
    "Derm7pt_clinical",
    "Derm7pt_derm",
    "MED-NODE",
    "MSK-1",
    "MSK-2",
    "MSK-3",
    "MSK-4",
    "PH2",
    "SDC-198",
    "UDA-1",
    "UDA-2",
    "ISBI2016",
    "ISBI2017_reduced",
    "ISBI2018_reduced",
    "ISBI2019",
    "ISBI2020",
]

for dataset in data_names:

    loaded = load_dataset(
        "z72pepee/dermatological-diagnoses",
        dataset,
        token="YOUR_TOKEN",
    )

    print("Dataset", dataset)
    print("Categories", loaded["train"].features["label"])
    print("Instances", loaded["train"].num_rows)

    print("Example 1", loaded["train"][0])

• Curated by: Eduardo Perez https://scholar.google.es/citations?user=csKAP9oAAAAJ&hl=en.
• Funded by: Research group KDIS (Knowledge Discovery and Intelligent Systems), University of Cordoba, Spain, https://www.uco.es/kdis/.