Efficient Dual-Encoder Network with Deep Dilated Convolutional Shuffle Sheaf Attention Network Fusion for Accurate and Robust Skin Lesion Segmentation and Classification

The Python code implements a hybrid deep learning pipeline for skin lesion classification, integrating AGGF-based image preprocessing, data augmentation, and dual-stage transfer learning feature extraction.Extracted features are fused using deep dilated convolutional shuffle sheaf attention blocks and refined through a CNN-based decoder for precise lesion segmentation.Final classification is performed using the Enhanced Hesitant Boundary-Aware Fuzzy Recurrent Neural Network with Transformer (EHFRNT), achieving highly robust and accurate lesion categorization.