<![CDATA[Acne vulgaris is one of the most common skin problems, particularly on the facial area, and it negatively affects the quality of life of sufferers both physically and mentally. The prevalence of acne continues to rise globally and nationally, especially among adolescents and young adults. This study aims to classify types of acne using the Convolutional Neural Network (CNN) method and to implement the results into an interactive web-based application using Streamlit, in order to facilitate users in independently detecting acne.The dataset used consists of 360 acne images collected from Google and Kaggle, which were categorized into four acne types: whiteheads, blackheads, pustules, and nodules before being split into training and testing datasets. This study employs three CNN architectures: InceptionV3, VGG16, and EfficientNetB0. Training was carried out in two stages with a learning rate of 0.0001 during the initial phase and 0.000005 during the fine-tuning phase, across a total of 50 epochs. The models were trained using the Adam optimizer, along with callbacks such as EarlyStopping, ModelCheckpoint, and ReduceLROnPlateau to prevent overfitting and enhance training efficiency.Model performance was evaluated using a confusion matrix. The evaluation results showed that the VGG16 architecture achieved the highest accuracy at 97%, followed by InceptionV3 with 96%, while EfficientNetB0 only reached 26% accuracy. The best-performing model was then integrated into a Streamlit-based application featuring a simple interface that allows users to upload facial images, detect acne types, and receive initial treatment recommendations.]]>
Copyrights © 2025
