Tire pattern classification based on ConvNeXt Network
Yuting Zhu, Mingkai Tao, Lingxi Guo, Xiaoqing Wang · 2022 IEEE International Conference on Visual Communications and Image Processing (VCIP) · 2022
Tire pattern classification is an important means to provide clues for traffic accident processing. With the rapid increase in the number of vehicles, it is in urgent need to develop efficient and automatic tire pattern image classification and recognition system. In this paper, we design a novel lightweight network model for tire pattern image classification dataset. The proposed method consists of two parts. On the one hand, the powerful ConvNeXt is used as the main baseline. On the other hand, the attention mechanism is introduced without sacrificing FLOPs, which improves the classification accuracy. Experimental results on the Tire pattern dataset demonstrate that, the proposed model performed satisfactorily when tested against low quality tire patterns taken from actual scenarios. In addition, experimental analysis of ablation demonstrates the good robustness of the proposed method.