Automated Fruit Classification using Enhanced Tunicate Swarm Algorithm with Fusion based Deep Learning

Abstract

Automatic fruit classification is an interesting problem in retail supply chain processes and in fruit industry, since the process can help the supermarkets as well as the fruit growers find various fruits and their statuses from the containers or stock. In this background, the current research article designs an Automated Fruit Classification system using Enhanced Tunicate Swarm Algorithm with Fusion-based Deep Learning (AFC-ETSAFDL) technique. The goal of the proposed AFC-ETSAFDL technique is to recognize different types of fruits from the captured images. In the presented technique, a fusion-based feature extraction method is applied using three DL models namely, DenseNet, ResNet, and Inception v3 with ETSA as a hyperparameter optimizer. For fruit classification, the Extreme Gradient Boosting (XGBoost) model is used in this study. The proposed model was experimentally evaluated using an open-access fruit dataset and the results showcase the enhanced performance of the proposed technique over other existing models.