CleanPatrick: A Benchmark for Image Data Cleaning

1. 1Collect a large dataset of images, such as the Fitzpatrick17k dermatology dataset
2. 2Annotate the images using a crowd workers platform to identify off-topic or noisy data
3. 3Use the CleanPatrick benchmark to evaluate the effectiveness of different data cleaning methods
4. 4Compare the performance of machine learning models trained on cleaned and uncleaned data
5. 5Apply data cleaning techniques to real-world image datasets to improve model robustness

Data scientists and machine learning engineers working with image data can benefit from using CleanPatrick to evaluate and improve their data cleaning pipelines

💡 CleanPatrick provides a large-scale benchmark for evaluating data cleaning methods in the image domain, enabling more accurate comparisons and real-world relevance

📸 Introducing CleanPatrick, a new benchmark for image data cleaning! 🚀 Improve your ML model's robustness with cleaner data 📊