Leveraging Data Augmentation and Dropout Layer in MobileNetV3 for Accurate Skin Cancer Detection ISIC Dataset

Skin cancer detection remains a critical challenge in medical diagnostics, requiring precise and efficient methodologies. This study aims to improve skin cancer detection using MobileNetV3, augmented with data augmentation and dropout layers, leveraging the ISIC dataset. Initially, the MobileNetV3 model achieved an accuracy of 91.24% with a loss of 25.05% when trained without data augmentation and dropout layers, highlighting the need for further optimization. To address these limitations, data augmentation was employed to increase the diversity of the training dataset, and a dropout layer was introduced to reduce overfitting. The enhanced model achieved a significantly improved accuracy of 95.83% and a reduced loss of 11.430/0. These findings underscore the model's robustness in accurately distinguishing between various skin lesion types, including actinic keratosis, basal cell carcinoma, melanoma, and others. This study not only demonstrates the efficacy of MobileNetV3 with advanced techniques in dermatological diagnostics but also sets a benchmark for future research in enhancing diagnostic accuracy through innovative model improvements. The successful integration of these methods in MobileNetV3 highlights its potential as a reliable tool for early and accurate skin cancer detection, ultimately contributing to improved patient outcomes.