Classification of Facial Skin Diseases Using Xception Convolutional Neural Network Model

There are many types of skin diseases on the face, including rashes that can be referred to as Acne, Eczema, and Rosacea. Additionally, the types of cells prominent in the epidermis can be referred to as keratosis and carcinoma. The overlapping descriptions of these skin diseases can be confusing, making it difficult to identify them based solely on their descriptions. This study explores the use of deep learning models CNN to improve the identification of skin diseases on the face. The dataset used is the Augmented Skin Condition Image Dataset obtained from Kaggle. By comparing the performance of different models, including Xception, ResNet50V2, DenseNet201, and the addition of the Adam optimizer, this research examines how different deep learning methods can enhance the identification of skin diseases on the face. The study finds that the selection of deep learning models can affect the performance of skin disease identification models. However, the use of the Adam optimizer does not significantly impact model performance. The optimal model results from using the Xception model, achieving 78% accuracy with the loss (0.1896) and accuracy (0.9495). For future research, you can apply feature engineering techniques and add hyperparameter optimization which can help in finding the optimal parameter combination.