A Comparative Study on Skin Cancer Detection Using Transfer Learning Models and CNN

Abstract

Dermatological diseases are the most common diseases in the world. Despite its prevalence, diagnosis is extremely difficult and necessitates extensive experience in the field. In this project, we present a method for detecting various types of these diseases. Computer vision and machine learning are two stages that we used to accurately identify diseases. The project's goal is to easily and accurately detect the type of skin disease and recommend the best treatment. The skin disease is subjected to various types of pre-processing techniques in the first stage of the image, followed by feature extraction. The second stage involves using machine learning algorithms to identify diseases based on skin analysis and observation. The proposed system is especially useful in rural areas where dermatologists are scarce. For the experimental results of this proposed system, we use a Pycharm-based Python script. Skin diseases are the most common around the world, as people get skin diseases due to genetics and environmental factors. In many cases, people ignore the early signs of skin disease. In the current system, skin diseases are identified through a biopsy process, which is then analyzed and medications are prescribed manually by physicians. We propose a hybrid approach combining computer vision and machine learning techniques to overcome this manual inspection and provide promising results in a short period of time. The input images for this would be microscopic images, such as histopathological images, from which features such as colour, shape, and texture would be extracted and fed into a convolutional neural network (CNN) for classification and disease identification. The project's goal is to easily detect the type of skin disease and recommend the best and most comprehensive medical recommendations. Skin disease much more rapidly and precisely. However, the cost of such a diagnosis is still limited and prohibitively expensive. Thus, image processing techniques aid in the development of an automated screening system for dermatology at an early stage. The extraction of features is critical in the classification of skin diseases. In a variety of techniques, computer vision plays a role in the detection of skin diseases. Skin diseases are common in Saudi Arabia due to the deserts and hot weather. This work advances the study of skin disease detection. We proposed a method for detecting skin diseases based on image processing.