<![CDATA[Accurate classification of coffee cherries based on ripeness is essential for enhancing the efficiency of harvesting and ensuring high-quality coffee production. Traditional manual sorting is labor-intensive and inconsistent, necessitating an automated solution. This study addresses the challenge by developing a mobile application that uses an object detection algorithm to classify Coffea canephore (Robusta) cherries into four ripeness categories: unripe, semi-ripe, ripe, and overripe. The application leverages a smartphone camera to capture images, which are then analyzed by a deep learning model trained on 1,200 annotated images, and classify coffee cherries in real-time. Model performance of the YOLOv5 computer vision was evaluated using a validation dataset (400 images) and a test dataset (400 images), ensuring balanced representation across ripeness levels. The application achieved an overall classification accuracy of 95.63%, with the highest accuracy for unripe cherries (98.50%), followed by semi-ripe (94.75%), ripe (94.75%), and overripe (94.50%) cherries. These results demonstrate the effectiveness of integrating mobile technology with object detection algorithm for field-based classification of coffee cherry ripeness. The developed application is potential for improving harvesting efficiency, optimizing quality control, and supporting decision-making in the coffee industry. Future work should focus on expanding the dataset, refining the classification model, and implementing the system in microcontrollers to enable an automatic sorting hardware, thereby reducing farmers’ workload and providing a comprehensive solution for our local stakeholders in the Bukidnon areas.]]>
