Improved MLP-Mixer for Cars' Type Recognition

With the development of intelligent manufacturing, the automotive industry, which is an important part of national economy, has attracted attention widely once again and the cars' type recognition is a crucial part for the automotive industry. The accuracy of cars' type recognition will directly affect the painting and other operations. Therefore, the car's type recognition requires a high accuracy over 99%. Some papers propose deep learning models for the cars' type recognition. However, many existing deep learning models have problems such as requirements for massive samples, slow convergence and difficulty in achieving the accuracy over 99%, which makes them have a little application to the industry. This paper takes the cars' type recognition as the background and adds the prior knowledge to the deep learning model, which introduces LBP into MLP-Mixer to improve the accuracy effectively.