Deep Learning-Based Exploration of YOLOv8 for Acne Vulgaris Type Classification and Lesion Counting

Acne vulgaris is a prevalent skin disorder that considerably affects teenagers and young adults worldwide. The condition is characterized by various lesion types, including papules, pustules, and nodules, each requiring specific treatment approaches. Conventional methods of acne assessment are often inconsistent, subjective, and labor-intensive, making it challenging for dermatologists to provide accurate and timely diagnoses. This study utilizes the You Only Look Once (YOLO) version 8 (YOLOv8) deep learning architecture to develop an automated and objective system for categorizing acne lesions and evaluating acne severity. Renowned for its speed and accuracy in object detection, this model processes facial images through resizing, normalization, and augmentation to maintain data consistency and boost detection precision. Lesions are categorized into papules, pustules, and nodules. The model achieved 80% precision, 81% recall, and a 78% Fl-score, illustrating its effectiveness in detecting and classifying acne lesions while minimizing diagnostic delays. Future efforts should address dataset imbalances by integrating underrepresented classes, exploring class-weighted training, and applying advanced augmentation techniques to further enhance performance. This study underscores the potential of AI-powered tools like YOLOv8 to advance dermatological practice, improve patient care, and contribute to public health progress.