Automated Grading of Bananas Using Advanced YOLO Models: A Machine Learning and Computer Vision Approach for Ripeness Classification

Among all fruits, bananas are consumed in high quantity around the world, and their health and industrial impact is enormous. The work presented here elaborates on how advanced machine learning and computer vision techniques can automate banana grading, particularly in its different versions of the YOLO model. In that respect, a dataset of 1899 images, representing five stages of ripeness, was manually created and pre-processed. Precision, recall, mean Average Precision at different IoU thresholds, and speed are applied to measure the performances of YOLOv8-n, YOLOv9-s, YOLOv10n, and YOLO11s. The experimental results demonstrated the high value of precision with 0.992 and a recall of 1.0 for the YOLOv8-n model, suitable for real-time applications, whereas YOLO11s attained the highest overall performance in various IoU thresholds with an mAP50-95 of 0.936. This work points out how automated grading systems can play an important role in improving both efficiency and the quality assessment of bananas, promoting consumer satisfaction while reducing post-harvest losses. Further research will be directed toward dataset enlargement, greater application of advanced imaging techniques, and overcoming practical challenges to implementation to improve the accuracy of grading and applicability in real situations, besides establishing a standardized framework for evaluation to gain more widespread acceptance of automated grading technologies in the agricultural sector.