SkinProNet: An attention-based deep learning system for skin disease classification and segmentation

SkinProNet is an AI-powered software tool designed for the classification and segmentation of skin lesions, including potentially life-threatening conditions like melanoma. It employs a novel hybrid deep learning architecture that combines advanced preprocessing methods with state-of-the-art models: EfficientNetV2Small for feature extraction, an optimized ACRNN for accurate classification, and U 2 -Net++ for precise lesion segmentation. This integrated approach enhances early detection and diagnosis of skin diseases. The model classifies six types of skin diseases with a high accuracy of 94.04% using both benchmark datasets and real-world clinical images. The results highlight the model’s potential as a reliable computer-aided diagnostic tool in dermatology. By leveraging attention-based mechanisms and efficient neural architectures, the software supports healthcare practitioners in diagnosing skin conditions quickly, accurately, and non-invasively. • A unified GUI-based pipeline for preprocessing, classification, and segmentation. • Combines EfficientNetV2, ACRN, and U 2 -Net++ for skin disease analysis. • Enhances image quality using Wiener filtering and contrast optimization. • Offers visual interpretability with feature maps and confidence scores. • Modular system design supports future integration and clinical deployment.