Hair Disease Detection with Deep Learning: A VGG16-Driven Model for Precise Classification of Hair and Scalp Conditions

This study enhances diagnostic precision by employing deep learning techniques to classify ten common hair and scalp disorders. A balanced dataset of 12,000 images from Kaggle was used with a 12,000 images total amount of 1200 images per class, and VGG16 convolutional neural network architecture was employed alongside transfer learning to learn intricate patterns present in medical images. Equal representation of all classes in training, validation, and test sets was ensured by partitioning the dataset into 80- 10 - 10 ratio of training, validation and test sets, respectively. The model was trained and evaluated on the Adam optimizer based on metrics such as accuracy, recall, precision, and F1 score. The model was able to achieve 98.24 % accuracy for the ten hair and scalp diseases and also traded above 97% precision and recall for all classes. This indicates that training and validation losses have a minimal difference between them, meaning that we have learned to generalize well without overfitting. By studying the disease confusion matrix, it turned out that the model could have a high true positive rate and relatively a lot of correct classification. The work presented here makes significant contribution in hair disease detection bringing deep learning models closer to seeing improved diagnostic accuracy in dermatology. Such models can not only help in early and precise diagnosis but can support telemedicine initiatives (especially critical during time where there is limited physical consultations). Other work includes expanding the dataset and exploring more advanced architectures to achieve even better model robustness and applicability to clinical settings.