Automated Visual Inspecting System for Fruit Quality Estimation Using Deep Learning

Abstract

Automated visual inspection using deep learning is widely used in recent years. In the field of agriculture, deep learning can be deployed to reduce effective man power, best time utilization, and supreme classification with improved accuracy. In agriculture, DL can be imported in many applications like soil identification, disease classification, fruit grading, and many more. Fruit quality classification is an essential part in farming as it implies to the return directly. Hence, an automated system is much needed to improve the classification of fruits with high accuracy and less time. In this paper, three different CNN models are proposed, namely simple CNN, ResNet50, and VGG19 for the said purpose. A comparison between CNN and two other forms of CNN (ResNet50 and VGG19) is presented in terms of performance. This work has achieved the optimum result with grading accuracy over 94% using VGG19 model.