C-net: a deep learning-based Jujube grading approach

Abstract

Jujube grading is a crucial process in the jujube-associated industry to ascertain the quality, ripeness, value, and security of the product. Traditionally, jujube grading has been done manually, which may be expensive, time-consuming, and prone to human mistakes. With the expansion of innovation, Machine Learning (ML)/Deep Learning (DL) turned out as a potent technique for automating the fruits grading process. Within this work, we deployed and analyzed the Concatenated-Convolutional Neural Network (C-Net) based on the residual network concept and seven cutting-edge CNNs for sorting the Indian jujube into six classes. To train and evaluate the models, we collected and assembled the dataset of jujube images. The performance analysis of the model relies upon two varying hyperparameters, batch size, and epochs as well as some performance metrics like F1-score, precision, and recall. The finding indicates that the proposed C-Net model was able to classify jujube images with high precision of 98.61% which surpasses other models but lags slightly behind the EfficientNet-B0 model. Our C-Net model has several advantages over most of the cutting-edge CNN models for jujube grading including increased accuracy, efficiency, cost-effectiveness, better decision-making, scalability, and real-time grading. The use of a C-Net model for jujube grading has the capability to revolutionize the jujube grading task and improve the fruit’s overall quality.