Modified YOLO network for symptom detection in panoramic oral roentgenogram

In this paper, a dental symptom detection model based on YOLO is proposed in order to detect different dental symptom in panoramic oral roentgenogram. This model introduces the Global Attention Mechanism into the backbone feature extraction network to obtain rich cross-latitude features and enhance the network’s global feature extraction capabilities in low-contrast images. At the same time, the Spatial Pyramid Pooling Fast module in the network is replaced and the Atrous Spatial Pyramid Pooling technology is used to improve the recognition ability of larger targets such as tooth germ. Finally, according to the special structure, size and position of different dental symptoms, the Focal-EIoU is introduced to replace CIoU, which increases the weight proportion of positive samples in the training process and reduces the problem of missed detection or false detection. Experiments on self-built data sets show that the improved YOLO model has improved mAP@0.5 by 4.3% compared to the original model, and the detection effect has been generally improved.