SkinDisNet: A comprehensive dataset of clinical images and metadata for skin disease detection and classification

Skin diseases are considered among the most important health problems worldwide, affecting millions of people. Early diagnosis can prevent complications and therefore improve the outcomes for patients. However, accurate diagnosis can be challenging due to the visual similarity of different conditions, variability in symptoms, and the shortage of region-specific annotated datasets for developing robust machine learning models. To bridge the gap, this paper presents a new dataset that aims to further the field of skin disease detection through machine learning and computer vision techniques. There are 1710 clinical images of skin diseases, captured using smartphones cameras from hospitals in Bangladesh. The images are organized into six classes of skin diseases-Atopic Dermatitis, Contact Dermatitis, Eczema, Scabies, Seborrheic Dermatitis, and Tinea Corporis. Each image is uniquely identified and stored in categorized folders, with all entries reviewed to ensure accurate classification. Moreover, this paper includes metadata that has been curated meticulously to provide essential clinical and demographic details, enabling precise analysis and reproducible research. Our dataset is intended to facilitate research and advancement in skin disease detection and classification using machine learning and computer vision techniques.