Real-Time Oil Palm Fruit Grading System Using Smartphone and Modified YOLOv4

The classification of the ripeness degree oil palm fruit has attracted attention numerous researchers. However, there are still many challenges due to constraints in dataset, methodologies used, and variations use data categories. Detecting bunches accurately is crucial, given their complex shape characteristics, particularly when different categories present a pile palm. Most studies utilize images or color spectrum classify level ripeness. these methods not real-time lack efficiency. This study proposes model for determining using smartphone video as input, incorporating modifications object detection approach. research process involves collecting videos piles smartphones grading area industry. then pre-processed labelled process. A developed YOLOv4 approach with several performance improvements, enabling implementation on smartphones. best-performing tested detecting classifying fresh an android-based smartphone. testing results, based mAP value, demonstrate that 16 quantization performs 12% better than Tiny. Based test results at location, this can efficiently detect do meet quality standards.