AI-Assisted Grading on Harumanis Mango

Abstract

In recent years, the increasing demand for fruits has presented significant challenges for fruit farmers and distributors, particularly in the realm of effective fruit management. A crucial aspect of this management is fruit grading, which is essential for assessing quality. Traditional manual grading methods, however, are prone to errors and inefficiencies, leading to inaccurate assessments of fruit maturity and quality. These inaccuracies cause substantial economic losses for distributors and hinder farmers' ability to deliver high-quality fruits to the market. This study focuses on the Harumanis mango (Mangifera indica Linn), which requires precise grading to meet market standards. By implementing an AI-assisted grading solution, distributors can efficiently grade Harumanis mangoes based on weight, size, and surface cleanliness, leading to significant reductions in labor costs and processing time. Our system's precision allows for strategic pricing according to grades, maximizing profitability. Testing of the system demonstrates its feasibility, with the classification model achieving 84% accuracy, the grading model 82% accuracy, and the weight estimation model reaching a low mean square error of 0.00184. These results highlight the potential of our AI-assisted grading solution to replace conventional manual methods, offering consistency and efficiency in fruit grading.