A modified YOLOv5 for object detection in UAV-captured scenarios

Object detection in UAV image processing has gradually become a hot research topic in recent years. The performance of general object detection algorithms tends to degrade significantly when applied to UAV scenes. This is due to the fact that UAV images are taken from high altitude with high resolution and a large proportion of small objects. In order to improve the precision of UAV object detection while satisfying the lightweight feature, we modify the YOLOv5s model. To address the small object detection problem, a prediction head is added to better retain small object feature information. The CBAM attention module is also integrated to better find attention regions in dense scenes. The original IOU-NMS is replaced by NWD-NMS in post-processing to alleviate the sensitivity of IOU to small objects. Experiments show that our method has good performance on the dataset Visdrone-2020, and the mAP is significantly improved from the original.