A Vehicle Detection Method Based on Improved YOLOv3

In order to improve the accuracy of vehicle detection, this paper proposes a vehicle detection method based on improved YOLOv3. First, the loss function of the YOLOv3 algorithm is improved by introducing the GIoU and focal loss strategies, meanwhile, the detection accuracy of the network is improved. Secondly, by adopting the method of data augmentation for preprocessing, adding label smoothing strategy improves the generalization ability of the model and further improves the average accuracy of detection. During training, multi-scale training is used, and the cosine decay learning rate decay strategy is used to replace the step learning rate decay strategy, which reduces the negative impact of improper learning rate and decay steps, speeds up model training, and reduces the network training cycle. Experimental results show that the improved method can significantly improve the detection accuracy of the original model in fewer training cycles, has good robustness, and can effectively detect road vehicles.