Analysis of Date Fruit Genetic Variety Based on Feature Selection and Classification Using Deep Learning Architectures

Deep learning has made significant strides recently, making it a popular option for image processing applications. However, because to similarities in color, shape, and size of fruits, researchers continue to have difficulty classifying them. This research proposes novel technique in date fruit genetic variety by feature selection and classification utilizing DL architecture. Here input is collected as date fruit images and processed for noise removal, normalization and smoothening. Processed input image is feature selected using convolutional radial basis network. Selected features are classified utilizing U-net transfer encoder learning architecture. For several date fruit datasets, experimental analysis has been done in terms of accuracy, precision, recall, F-1 score, RMSE, and mean precision. Proposed technique attained accuracy of 97%, precision of 93%, recall of 83%, F-1 score of 89%, RMSE of 63% and mean precision of 83%.