Comparative Study of CNN Models for Defect Detection in Food Packets

Abstract

Industry 4.0 is the term which promises a new industrial revolution. It is an amalgamation of advanced manufacturing techniques and Internet of Things(IoT) to produce such manufacturing systems which are interconnected, and can communicate, do analysis, and utilize the information to drive further intelligent action back in the physical world. Industrial Internet of Things (IIoT) involve application of IoT in manufacturing and other industrial processes to enhancing the working condition, and improvement of operational efficiency (Foukalas et al.). This paper reviews the recent work on industry 4.0 for automated defect detection in food packaging industry. This will help to reduce the complexity and improve the speed and accuracy of detection. This paper discusses the challenges and applications of industry 4.0 in general and further proposes a method to compare how various CNN models can be used for detecting the defects in food packaging industry. In this work seven (Alexnet, Resnet50, Resnet101, Densenet, VGG16, VGG19 and Squeezenet ) different convolution neural networks are subjected to detecting the defects in food packets. After running the models with a Multi-Label-Classifier the training accuracy after 100 epochs is found as 98.5% and Validation Accuracy as 98.4%