Hybridizing Convolutional Neural Networks and Support Vector Machines for Mango Ripeness Classification

Abstract

This research aims to identify the maturity of mangoes by proposing a hybrid approach that combines a convolutional neural network (CNN) and support vector machine (SVM). Sorting mangoes according to ripeness is a vital agricultural exercise that increases yield productivity and reduces overages during storage. The suggested hybrid model aims to improve the efficiency and accuracy of existing methods for classifying mango ripeness. The hybrid CNN-SVM model was trained and tested using the dataset containing approx. thousand images of mangoes in three stages (unripe, ripe and overripe). The proposed hybrid method combines CNN's capability to extract characteristics from visual input with the accuracy of SVM classification. With a farfetched 98.53% accuracy rate, experiments with the hybrid model show that it performs better than both traditional machine learning and deep learning approaches. These results demonstrate how hybrid models may be used to assess the maturity of mangos quickly and accurately, which might improve agricultural decision-making.