Small Target Detection Algorithm Based On Multi-target Detection Head And Attention Mechanism

Abstract

To solve the problem of low detection precision and accuracy of small and medium-sized targets in aerial UAV images, an improved YOLOv5 algorithm is proposed. Fine-grained feature detection of small targets is realized by fusing Depthwise Separable Convolution multi-target detection heads in feature enhancement networks. The CBAM attention mechanism module is embedded in the backbone network to reorganize different feature channels, and assign large weights to key features to highlight the semantic information and detailed features of small targets, effectively weaken background interference, and improve the algorithm's overall perception ability and detection accuracy. Experiments were carried out on the UAV aerial photography data set VisDrone2019, and the results show that the average accuracy of the improved algorithm is increased by 5.5% and the accuracy is increased by 4.2%, which has a better detection effect.