SKIN ACNE SKIN DISEASE CLASSIFICATION BY USING FINE TUNED CONVOLUTIONAL NEURAL NETWORK

Abstract

Acne vulgaris is a prevalent and chronic inflammatory skin condition, categorized by blackheads, whiteheads, pimples, nodules, and cysts, affecting up to 80% of adolescents and often extending into adulthood. The psychological, social, and economic effect of acne is profound, contributing to issues such as low self-esteem, depression, and in severe cases, suicidal thoughts. Traditional methods for assessing acne rely on the manual expertise of dermatologists, which, while valuable, can be time-consuming, prone to observer inconsistency, and potentially partial. Accurate and timely diagnosis is essential for effective treatment, making the need for automated acne classification systems increasingly apparent. Advances in artificial intelligence, particularly Convolutional Neural Networks (CNNs), present promising opportunities to enhance the accuracy and consistency of acne classification. This study develops acne types classification system using a fine-tuned CNN model and evaluate its performance against well-established pretrained models, including ResNet50, Mobile NetV3, and InceptionV3. The proposed model achieved an accuracy of 71% (proposed model), ResNet50 (62%), MobileNetV3 (49%), and InceptionV3 (46%). By bench-marking the proposed model’s effectiveness against these pre-trained architectures and existing research, this research demonstrates improvements in diagnostic accuracy, reliability, and efficiency. Implementing such automated systems has the potential to enhance treatment outcomes, reduce the burden on healthcare professionals, and provide a more consistent and objective approach to acne diagnosis.