Implementation of Convolutional Neural Network (CNN) in Android-Based Acne Detection Applications
Tutiani, Sutresna Wati · 2022
Acne is a skin disorder caused by the overproduction of oil glands that causes infection and inflammation of the skin and is one of the most common conditions in the world. The Convolutional Neural Network (CNN) approach method of convolution operation is used to detect objects in the assessment of acne and the correct handling solutions to overcome them in the form of an android-based application. Data processing is carried out in several stages. The initial phase of data processing use LabeIImg software to label the image object, producing a file with an XML extension. The Efficientdet-lite model is a model that has been optimized to run on mobile devices with the Efficientdet-lite0 model architecture in the modeling process that has been trained using the hyperparameter model, which has a detection loss of 0.056 and produces an Average Precision value of 0.1256. The androidbased acne detection application was successfully created using the CNN method and the Efficientdet-lite model, the application work on smartphones with a minimum version of nougat 7.1. The results of the Average Precision values from the labeling of seven classes are 0.0627, 0.0027, 0.1575, 0.0217, 0.0038, 0.01224, and 0.014, so it can be concluded that the average precision value in each category is quite good but must be developed to achieve more accurate detection performance.