Performance Comparison of VGG16, MobileNetV2, and InceptionV3 Convolutional Neural Networks in Classifying Facial Dermatological Conditions

This study investigates the performance of three convolutional neural network (CNN) architectures (VGG16, MobileNetV2 and InceptionV3) in classifying two common facial dermatological conditions: acne and dark spots. A dataset of 235 facial skin images was augmented, then used to train and evaluate each model using standard classification metrics such as accuracy, precision, recall, and F1-score. The results demonstrate that MobileNetV2 achieved the highest classification accuracy of 93.13% while maintaining a relatively low computational cost. The model exhibited perfect precision (1.00) for the acne class and a high recall of 0.99 for the dark spots class, indicating its strong capability in accurately and sensitively identifying both lesion types. All three models demonstrated acceptable classification performance for both acne and dark spots classes, as evidenced by their precision, recall, and F1-scores exceeding 70%. This indicates that each model was capable of capturing relevant discriminative features of both lesion types.