Benchmarking Models with GRAD-CAM Visualizations

Skin cancer is one of the most common and potentially deadly forms of cancer worldwide, making early and accurate detection crucial for effective treatment. This project presents a comprehensive benchmarking of five deep learning models to classify skin cancer images and identify the most accurate architecture for diagnosis. The system utilizes a publicly available dermatoscopic image dataset and implements advanced convolutional neural network architectures, including VGG16, ResNet50, InceptionV3, MobileNetV2, and EfficientNetB0. Each model is trained and evaluated based on standard performance metrics such as accuracy, precision, recall, and F1-score. The models undergo uniform preprocessing and training conditions to ensure fair comparison. Results demonstrate that certain models consistently outperform others in terms of classification accuracy and computational efficiency. This benchmarking study provides valuable insights for medical AI developers and healthcare practitioners by identifying the most effective deep learning model for skin cancer detection. The findings emphasize the potential of deep learning in enhancing diagnostic accuracy and enabling scalable, automated solutions for dermatological analysis.