Currently, dermatologists need to check numerous image reports (high resolution) for diagnosing skin conditions, and Machine Learning (ML) models can help with this tedious task. However, current ML models usually work best with high-quality images in case of skin cancer, and those high-resolution images need dermoscope. In this paper, we investigated how well ML performs on low-quality ones. Our results shows though there are challenges with using lower-quality images but still models perform with >84% accuracy, and we plan to improve the models to make them more accurate and useful in real-world clinics.
Keywords: Dermatology; Image Classification; Low Resolution; Skin Cancer.