A CNN-Based Hybrid Approach to Classification of Raisin Grains

Abstract

Raisin Grains, which are an important food source thanks to their rich carbohydrates, potassium and iron contents, are also beneficial for many health problems. When the classification of the type and quality of raisin grains is done with traditional methods, it can be easily affected by the psychological and physiological condition of the specialist who performs the operation. For this reason, it is important to realize systems based on machine learning methods in order to obtain more successful and reliable results. In this study, we focused on CNN-based hybrid machine learning methods for the classification of 2 different types of raisin grains. Evaluations were made using 5 different machine learning methods: KNN, Ridge Classifier, XGBoost, SVC and LDA. In order to evaluate the CNN-based hybrid model, raisin grains were first classified by the classical method using these machine learning methods. Then, classification operations were performed using CNN + Machine Learning methods and compared with the results obtained with classical machine learning. As a result of the study, when the results obtained with the hybrid model proposed in the study were compared with the results obtained with the classical methods, it was seen that the hybrid model increased the success compared to the classical machine learning methods.