Deep learning-based skin disease detection

Abstract

Skin diseases are more common than other disease in Bangladesh. Skin diseases may be caused by fungal infection, bacteria, hot weather, allergy, or viruses, etc. The thesis paper on skin disease detection focuses on the application of advanced technologies, specifically artificial intelligence and machine learning, for the early and accurate identification of various skin conditions. The research involves the acquisition of skin images through diverse sources such as smartphones, cameras, or specialized devices. These images undergo a systematic process, including image preprocessing, feature extraction, and classification, facilitated by sophisticated algorithms. There are various types of algorithm which using in disease detection method. This algorithm is, Convolutional Neural Networks (CNN), Support Vector Machines (SVM), Random Forest (RF), Transfer Learning (Pre-trained CNNs), K-Nearest Neighbors (KNN), Feature Extraction Techniques (e.g., HOG, GLCM, LBP) and Recurrent Neural Networks (RNNs). Determining the "best" algorithm for skin disease detection depends on various factors, including the nature of the dataset, the specific skin conditions targeted, and the characteristics of the images involved. Normally we decide to use CNN algorithm because of, CNNs offer many advantages for skin disease detection, it's essential to consider the specific requirements and constraints of the application when choosing an algorithm. The proposed CNN architecture is based on VGG-16 and will be trained and tested on datasets collected from skin disease. The goal is to create a reliable and non-invasive method for diagnosing skin conditions, with potential applications in both self-assessment tools for users and remote consultations with healthcare professionals.