Deep learning-based detection of viral skin diseases using ResNet-152 with web deployment

Abstract

Viral skin infections remain a significant public health concern, particularly in resource-limited regions where access to dermatological expertise is constrained. This study presents a deep learning–based system for automated detection of viral and related skin diseases using image classification and segmentation techniques. A ResNet-152 convolutional neural network was fine-tuned using the FastAI framework and trained on an augmented dataset of 3,028 images derived from 703 original samples across five conditions: monkeypox, chickenpox, warts, eczema, and corns. Data preprocessing and augmentation techniques were applied to address class imbalance and improve generalization. The proposed model achieved an overall classification accuracy of 92%, with notable improvements observed in underrepresented classes such as eczema after augmentation. Unlike prior studies primarily focused on skin cancer or single-disease classification, this work emphasizes viral skin infections and integrates the trained model into a real-time Flask-based web application for practical deployment. The results demonstrate the effectiveness of deep residual networks in multi-class dermatological classification and highlight the potential of AI-driven tools for accessible and early skin disease screening.