Implementation Of Convolutional Neural Network (CNN) in The Android-Based Application for Detecting Coffee Bean Maturity

The research aims to explore the implementation of artificial intelligence in developing an application that can accurately detect the maturity level of roasted coffee beans solely using a smartphone device. The research employs an AI-based approach by utilizing deep learning techniques to train a model on a dataset containing various roasted coffee bean maturity levels. The training data comprises visual images of coffee beans at different stages of maturity, enabling the model to discern patterns and characteristics associated with maturity levels. The application is built upon a classification system model for the maturity level of roasted coffee beans, employing an algorithm or convolutional neural network (CNN) method using the ResNet50 architecture. The utilization of the CNN method can enhance the accuracy of the model. This system is developed by employing digital image processing with data represented as coffee bean images. The smartphone application system uses Visual Studio Code, Flutter, and Android Studio software, incorporating image processing, colour feature extraction, and CNN classification methods. The model evaluation demonstrates an accuracy level of 100%, with precision, recall, and F1-Score values also reaching 100%. However, the accuracy of the ResNet50 model in classifying the maturity level of roasted coffee beans in real-time applications, based on the confusion matrix, is approximately 83.3%.