Machine Learning-based Automatic Grading of Mangoes
Juwel Das Asish
Bangladesh Agricultural University, Mymensingh, Bangladesh.
Rakib Hassan
Bangladesh Agricultural University, Mymensingh, Bangladesh.
S M Abdullah Al Shuaeb *
Bangladesh Agricultural University, Mymensingh, Bangladesh.
Md. Mizanur Rahman
Jatiya Kabi Kazi Nazrul Islam University, Trishal, Bangladesh.
*Author to whom correspondence should be addressed.
Abstract
Automation in agriculture can improve product quality and consistency while reducing labor and time costs. Manual mango grading is often slow, subjective, and error-prone; therefore, an automated grading approach is desirable. In this study, we propose an image-based mango grading framework that classifies mangoes into four grades (green, semi-ripe, ripe, and rotten) using five models: Artificial Neural Network (ANN), Decision Tree (DT), k-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Convolutional Neural Network (CNN). We collected 4,207 mango images from multiple markets and locations and used a 70%/30% train–test split. Performance was evaluated using Accuracy, Precision, Recall, and F1-score across five repeated runs. CNN achieved the best overall performance (mean accuracy ≈ 0.99) and consistently outperformed the classical machine-learning baselines, indicating that deep learning can provide a reliable solution for automated mango grading and can be extended to other agricultural products.
Keywords: Mango grading, Computer vision, image classification, machine learning, deep learning, Convolutional Neural Network (CNN), fruit quality assessment