Classification of Dry Beans into Genetic Varieties Using Deep Learning-Based Convolutional Neural Networks (CNNs)

Efficient crop management, breeding programs, and the maintenance of sustainable agriculture are all dependent on the critical task of identifying and classifying genetic variations in crops. The study entailed collecting and preprocessing an extensive dataset comprising morphological, physical, and genetic attributes of different dry bean samples. The primary goal was to develop a robust classification model that could discern the genetic variations among various dry bean varieties. To ensure consistency, a preliminary phase of data preprocessing was executed, including color normalization and data augmentation. The findings demonstrated the effectiveness of the proposed deep learning-based method, achieving high accuracy and reliable classification of dry beans, including Seker, Barbunya, Bombay, Cali, Dermosan, Horoz, and Sira into their genetic variations. To accomplish this goal, supervised deep learning techniques have been subjected to rigorous training using one distinct dataset: one containing 70% training data and 30% testing data. These processes were implemented to acquire knowledge and classify the intricate patterns that are inherent in the given data. The classification of Custom CNN is incredibly accurate at 99.85%, while the Xception and MobileNet accuracy in classifying dry bean samples is slightly lower at 82%.