Enhanced Skin Cancer Classification with AlexNet and Transfer Learning

The rising incidence of skin cancer presents a growing public health challenge, underscoring the critical imperative for the development of efficient and precise detection and classification methodologies. Patients grappling with the prospect of a skin cancer diagnosis often find themselves embroiled in prolonged diagnostic processes, leading to delayed treatment initiation and potentially compromising their chances of survival. In light of these challenges, the strategic application of digital image processing assumes paramount importance. Within this context, the extraction of salient features from dermatological images emerges as a pivotal step in the accurate identification of various skin cancer types. Convolutional Neural Networks (CNNs) is known for its ability to autonomously extract relevant features from complex images, offer a promising solution for achieving high-accuracy skin cancer diagnosis. This research study showcases the capabilities of a CNN model based on the AlexNet architecture, which is further enhanced with advanced Transfer Learning (TL) techniques. This research study utilized the dataset sourced from the Human Against Machine (HAM 10000) repository, encompassing two distinct cancer types and a non-cancer category for comprehensive analysis.