Recognition of expiry data on food packages based on deep learning

There is a great deal of information on food packaging, including the name of the food, the expiry date and the ingredients. This information, especially the expiration date, needs to be printed correctly before the product is brought to market. Failure to print the correct expiration date can lead to public health problems and recalled products causing financial losses to the company. In this work, we propose an automatic detection and identification of validity areas that can be achieved efficiently and accurately. First, the DBNet network-based approach is applied to detect the expiration date region on food packages. Then the detected expiration date area is intercepted and input to the character recognition network CRNN for character recognition. Finally, the proposed model is deployed on Jetson Xavier NX to implement edge computing of the algorithm, while inference acceleration of the model is performed using TensorRT and FP16 or INT8 quantization. The experimental results show that the method achieves good performance in the detection and identification of expiration dates on different types of food packages, and the method has good real-time and portability.