Yang, Dayeon (Chonnam National Univ.), Ju, Chanyoung (Korea Institute of Industrial Technology)

Performance Comparison of Cherry Tomato Ripeness Detection Using Multiple YOLO Models

Scheduled for presentation during the Regular Session "AI-based Robot Control II" (WeBT2), Wednesday, July 16, 2025, 15:20−15:40, Room 106

Joint 10th IFAC Symposium on Mechatronic Systems and 14th Symposium on Robotics, July 15-18, 2025, Paris, France

This information is tentative and subject to change. Compiled on July 16, 2025

Abstract

Millions of tons of cherry tomatoes are produced annually, highlighting the urgent need for efficient harvesting systems to sustain agricultural productivity and meet global food demands. This paper presents a deep learning-based solution for real-time ripeness classification, focusing on YOLO (You Only Look Once) v5 and YOLOv8 models. The YOLOv8 model was further enhanced with a ResNet50 backbone to improve feature extraction and classification performance. Using a custom dataset of 742 images, expanded from an initial 300 through data augmentation techniques such as rotation, brightness adjustment, and flipping, YOLOv8 with ResNet50 achieved a mean average precision (mAP) of 0.757, surpassing YOLOv5 by 5.6%. This model not only ensures higher detection accuracy but also enables efficient real-time processing essential for agricultural automation. By addressing labor shortages and improving operational efficiency, this approach lays the groundwork for intelligent harvesting robots equipped with advanced segmentation capabilities and automated manipulators. These innovations promise to revolutionize precision agriculture by enabling faster, more accurate cherry tomato harvesting in diverse environmental conditions.