Ripeness Detection and Robotic Automation: A Comparison of Deep Learning Models

Contemporary agricultural practices have recently used deep learning algorithms and robotic technologies to improve operational efficiency, particularly in the case of assessing ‘fruit ripeness’. The objective of this study is to compare deep learning models used to detect the ripeness of bananas and mangoes with high accuracy. For real-time processing, these models are then integrated into robotic systems for ‘fruit picking’. Four deep learning models, namely Custom CNN, MobileNetV2, Xception, and YOLOv8, are presented in this chapter for image classification and object detection tasks. The experimental results showed that Xception and MobileNetV2 have higher accuracy rates in the ripeness classification. Results show that YOLOv8 has strong capabilities for both localization in real time and overall picking accuracy during simulations of 88%.