Intelligent fruit quality classification system using transfer learning

Abstract

<p>Amidst the burgeoning demands of fruit agriculturists and grading companies for enhanced fruit quality classification, this research presents a cutting-edge approach to binary fruit quality assessment. We built a portable device for exact fruit quality inspection using transfer learning, a deep learning approach, resulting in a decrease in both human and machine labor. The performance of the system is validated and evaluated under real-time situations, with an emphasis on end-user applicability. This paper rigorously validates and assesses the system’s performance in real-world scenarios, with a strong focus on its practicality for end-users. The model is trained on an online picture dataset that is divided into two categories: ‘good’ and ‘poor’ fruits. On dataset 1, our numerical findings show outstanding classification accuracies of 99.49% and 99.75% for the first and second models, respectively. Meanwhile, on dataset 2, the first and second models attain accuracies of 85.43% and 96.75%, respectively, highlighting the efficacy of our technique.</p>