Multi-Model Attentional Fusion Ensemble for Accurate Skin Cancer Classification

◦ An ensemble combining ResNet50V2, MobileNetV2, and EfficientNetV2 with attention mechanism to improve skin
cancer classification accuracy.
◦ Achieved 92% precision and 96% recall for Dermatofibroma, and 99% precision and recall for Vascular lesions,
demonstrating superior results across varied skin lesion types.
◦ Utilized resampling, augmentation, and hair removal techniques to address class imbalance in the HAM10000
dataset, enhancing overall model performance.