Deep Learning Approach for Palm Fruit Ripeness Classification Using MobileNet

Palm oil is a critical global commodity, with its quality largely determined by the ripeness of the palm fruit at harvest. However, improper harvesting practices often lead to issues such as harvesting unripe or rotten fruit, reducing oil quality and yield. This study aims to develop an automated solution for palm fruit ripeness detection using deep learning, specifically employing the MobileNet model on edge computing devices to address the challenges faced in palm oil plantations. A descriptive qualitative research approach was adopted, where the MobileNet deep learning model was trained and embedded on an Android phone to classify palm fruit ripeness. The model’s performance was evaluated in real-world plantation conditions, yielding an accuracy of 85%. The findings highlight the potential of utilizing edge computing for palm fruit ripeness classification, offering a practical solution to reduce human error and improve harvesting efficiency. This study demonstrates the feasibility of implementing deep learning models in resource-constrained environments, emphasizing the significance of integrating AI technologies in agricultural practices. The results indicate that the system could enhance the quality of palm oil production by ensuring more accurate and timely harvesting, thereby contributing to more sustainable and efficient palm oil supply chains. Future work may focus on optimizing the model for real-time applications and extending the system’s capabilities for broader agricultural use cases.