<![CDATA[Accurate classification of acne types is essential for determining the appropriate treatment. This study developed an automatic detection system based on Android using an ensemble Deep Learning approach and partial layer optimization to address the visual similarity challenges among different acne types. MobileNet and EfficientNetB0 were used as base models, then combined using majority voting and weighted averaging techniques. The dataset used consisted of 6,875 acne images that underwent preprocessing and augmentation. To improve efficiency and prevent overfitting, early layers of the models were frozen, and fine-tuning was applied only to the top layers. The best-performing model was then converted into TensorFlow Lite (TFLite) format and integrated into an Android application. The application allows users to classify acne in real time using either the camera or gallery. Evaluation results showed that the ensemble model offered better accuracy and stability compared to individual models, with fast inference times on Android devices. This system provides a practical and accurate solution for both general users and medical professionals in detecting acne types.]]>
Copyrights © 2025
