Automated Sorting of Date Fruits: Integrating Deep Learning and Discriminant Correlation Analysis

This paper suggests a new method for classifying various types of date fruit. It combines supervised and unsupervised deep learning approaches. The main idea behind this approach is to automate the sorting process, which usually depends on manual labour. It is very costly and time- consuming. By using deep neural networks - particularly the VGG-F convolutional neural network and a network called PCANet - this new method aims to simplify the classification process. Discriminant Correlation Analysis (DCA) is used to blend features learned from both networks. It also helps in reducing the dimensionality, but with low computational complexity. This blend of supervised and unsupervised techniques allows for a better understanding of the data, whichcould lead to enhanced accuracy in classification. To assess the efficacy of the new method, a fresh benchmark dataset is introduced. This dataset consists of images of date fruit from around 20 different varieties. The dataset stands out due to its size and diversity, making it a valuable resource for the research community. It is also shared publicly to aid further research and bench marking. The experimental outcomes show the success of the DCA approach and how the fused features complement each other. Moreover, the proposed method surpasses various relevant methods, hence proving its efficiency in date fruit classification. In conclusion, this paper presents a unique method for classifying date fruit. It utilizes both supervised and unsupervised deep learning techniques, showing promising results through experiments on a vast and diverse dataset.