Image Data for Beef and Lamb Slices Authentication

Meat adulteration affects customers and the market. Existing meat authentication methods usually relay on special devices, and thus are limited to professional use only. Based on deep learning technique, we present a customer-manipulated method for hotpot lamb and beef slices authentication. This method takes a single image as input, and obtain authentication result rapidly. We build a lightweight (thus high-efficiency) convolutional neural network architecture called MTx-Net for this task. For training and testing the neural network, we collected 77956 meat images using over 225kg of meat. Our method achieves 99.38% and 98.20% accuracy on lamb and beef slices, respectively.Image data, code and trained model are uploaded here. The image data include genuine and fake lamb slices and beef slices. The fake meat slices are made of duck and lamb or beef fat. A lite subset of meat images is also provided as "Image Examples.rar", which envolves 60 images.