Exploring and Classifying Beef Retail Cuts Using Transfer Learning

Abstract

An evaluation of the deep learning neural network in artificial intelligence (AI) technologies is proposed to provide a rapid recognition and immediate proper classification of the different beef retail cuts (Liver, Roast Beef, Beef Chuck, Beef Round, Strip-Lion, Round Fillet, Beef Flank) to classify them accordingly. The problem is that many of the modern consumers face difficulties in recognizing the different retail beef cuts. Thus, a solution was created through collecting a dataset for retail cuts and creating an algorithm to classify them. A dataset, which is available for public, of 7 different beef retail cuts was proposed. This dataset includes colored images from our own image library, a total of 1638 images for validation testing and training are used for this project. The deep learning neural network algorithm-based model was able to identify specific beef retail cuts. 5 models were used in this paper to reach the highest accuracy for the classification of our dataset (MobileNet, ResNet50, InceptionV3, EfficientNetB0 and our customized model). EffecientNetB0 pretrained model is one of the best and easiest pretrained models in Keras CNN. The employment of this model, after training and data augmentation techniques, was able to achieve the highest accuracy by a 99.81%. Based on our trained model and the huge results, deep learning technology evidently showed a promising effort for beef cuts recognition in the meat science industry.