YOLO, Faster R-CNN, and SSD for cloud detection

Abstract

Object detection is an essential problem in computer vision. Many models perform well in different kinds of object detection problems. However, there needs to be more research on object detection for similar objects in traditional research fields. To study the performance of standard target detection models in similar target detection, this paper uses the cloud detection problem that requires higher accuracy than detection speed as an example. This paper trained and tested three models, YOLO, Faster R-CNN, and SSD, with our data set and obtained excellent detection results. On this basis, this paper puts forward some reasons that may cause the detection accuracy not to be high and puts forward the corresponding optimisation methods for these reasons, hoping to provide some ideas and help to solve this kind of problem.