Enhancing Skin Cancer Detection with Novel Data Augmentation and Transfer Learning Techniques

Skin cancer is one of the most prevalent and life-threatening diseases worldwide, requiring early detection for effective treatment. This study investigates a deep learning-based approach for classifying skin cancer into multiple categories, utilizing CNNs and transfer learning models. Initially, a CNN model with reduced complexity was applied to classify skin cancer images from a dataset containing various classes, including melanoma, basal cell carcinoma, and nevus. Due to the limited size of the dataset, data augmentation techniques were used to generate additional samples for each class, improving model performance. Transformations such as rotation, zoom, shifting, and flipping were applied to simulate real-world variability. To further enhance classification accuracy, four pre-trained transfer learning models—VGG16, ResNet50, InceptionV3, and EfficientNet—were fine-tuned and compared. The results shown a substantial improvement in accuracy with the combination of data augmentation and transfer learning.