<![CDATA[The quality of Crude Palm Oil (CPO) is highly dependent on the accuracy of sorting the ripeness level of oil palm Fresh Fruit Bunches (FFB). Manual sorting processes currently used in factories are vulnerable to human error and subjectivity. This study aims to automate the objectivity of the sorting process using a deep learning model based on the ResNet-50 architecture with a transfer learning approach to classify FFB into three categories: Unripe, Ripe, and Overripe. The computational model was integrated into a web-based application using the Flask framework to support wireless operational use in factories. Experimental results showed a validation accuracy of 90.94% and an F1-score of 91%. Direct field validation using 42 primary data samples achieved a classification success rate of 83.33%. The implementation of a 75% confidence threshold proved effective in preventing prediction errors (zero misclassification), while the Cohen’s Kappa reliability test achieved a score of 0.769, indicating Substantial Agreement with expert evaluators. In conclusion, the ResNet-50-based system demonstrated reliable and objective performance and is considered ready for replication to maintain quality consistency in the palm oil processing industry.]]>
Copyrights © 2026
