Vehicle Logo Detection Method Based on Improved YOLOv4

A vehicle logo occupies a small proportion of car and has different shapes. These characteristics bring difficulties to machine-vision-based detection. To improve the accuracy detection in complex backgrounds, an improved YOLOv4 model was presented. Firstly, CSPDenseNet introduced backbone feature extraction network, shallow output layer added replenish information target. Then, deformable convolution residual block employed reconstruct neck structure capture various irregular shape features. Finally, new head based on convolutional transformer proposed reduce influence backgrounds Experimental results showed that average all categories VLD-45 dataset 62.94%, which 5.72% higher than original model. It indicated could perform well