Deep Learning-Powered Hair Disease Diagnosis: A ResNet50 Approach for Scalable and Accurate Classification

This study introduces a deep learning framework using the ResNet50 model for automated classification of hair diseases, achieving a 95% accuracy on a dataset of 10 categories with 9600 training and 1200 validation images. The model demonstrated high true positive rates for conditions like Alopecia Areata and Lichen Planus, utilizing metrics such as precision, recall, and F1 score. Techniques like data augmentation and learning rate adjustments were employed to enhance generalization. Despite some misclassifications between similar conditions, the model shows promise for scalable, accurate, and efficient dermatological diagnostics. Future directions include expanding the dataset and incorporating interpretability tools like Grad-CAM.