Skin diseases detection using deep learning techniques with web application

Abstract

A widely recognized paradigm exists for imaging purposes in the identification of skin disorders. In recent years, many medical professionals, including physicians, use CAD to assist them more accurately detect a range of disorders by looking at clinical pictures. Skin cancer is one of the most deadly diseases in the world. It's challenging to diagnose skin cancer correctly, though. This research offers several image processing techniques to help diagnose different kinds of skin diseases. The purpose for this research is to ascertain whether or not both these eights’ skins are diseases, as well as to identify the type of skin illness by utilizing a variety of skin disease classes. In addition to eight additional class types, deep learning-based methods were employed in this experiment: 'BA-cellulitis', 'BA-impetigo', 'FU-athlete-foot', 'FU-nail-fungus', 'FU-ringworm', 'PA-cutaneous-larva-migrans', 'VI-chickenpox', and 'VI-shingles'. Five different models are used: DenseNet169, InceptionV3, VGG16, VGG19, and InceptionResNetV2 to forecast and detect skin photos and categorize illnesses. Lastly, two distinct efficiency metrics are used to evaluate the approach's performance. Five possible results are used in the initial accuracy set, which rates performance under both normal and fractured circumstances: TP, TN, FP, and FN. These models are then applied to investigate the exact nature of each sort of illnesses in mistake settings. With an accuracy percentage of 98.23%, the InceptionResNetV2 technology enables my suggested method to independently identify different kinds of skin disorders. In the end, data classification using the InceptionResNetV2 networks to identify skin illnesses results in the creation of an application in web.