Digital-Dermatology/CleanPatrick

CleanPatrick: A Benchmark for Data Cleaning Welcome to CleanPatrick, the first large-scale benchmark designed for data cleaning in the image domain. Built on the Fitzpatrick17k dermatology dataset, CleanPatrick is a dataset for measuring the performance in detecting three major data quality issues: off-topic samples, near-duplicates, and label errors. Overview CleanPatrick consists of dermatological images annotated with over 500,000 binary labels across three… See the full description on the dataset page: https://huggingface.co/datasets/Digital-Dermatology/CleanPatrick.