Tire Radiogram Defect Detection Based on ESSNet

To address the issue that tire defects cannot be accurately localized in tire defect detection, an intelligent tire defect detection model based on simple semantic segmentation (ESSNet) is studied, which consists of a simple encoder-decoder network and a multi-scale residual refinement module, responsible for anticipating the tire X-ray coarse feature map and refining the coarse feature map mapping, respectively. Additionally, a hybrid loss function is employed to achieve more precise localization limits for the object to be recognized, and a texture segmentation approach is used to minimize the complexity of the tire X-ray image. It is demonstrated that the model is capable of precisely locating and segmenting the defects in the tire pictures and achieving a mPA score of 86.24 % on the test set.