This project implements a state-of-the-art skin cancer classification system that combines Convolutional Neural Networks (EfficientNetV2, ConvNeXtV2) with Vision Transformers (ViT, Swin Transformer) through an adaptive attention-based fusion mechanism. The system classifies dermoscopic images into 7 skin lesion categories from the HAM10000 dataset.