Drone-View Object Detection Based on the Improved YOLOv5

With the rapid development of computer vision and machine vision, deep learning-based methods have achieved good results in the fields of object detection, identification, and tracking. However, for the detection and identification of vehicles and pedestrians under drone views, the drone-view object detection algorithm detects poorly because the features of small size targets are difficult to be extracted and identified. In order to improve the accuracy and performance of drone view object detection, this paper proposes an improved YOLOv5 algorithm that adds a feature extraction layer to the network model and an image segmentation layer to the detection network. Specifically, a feature map for collecting small target features is formed by adding one upsampling to the neck portion of the network, and a new feature extraction layer is formed by concat. Experimental results show that the proposed method improves the detection effect of the YOLOV5 algorithm on small targets.