UAV-ROD is a car detection dataset for UAV and annotation with oriented bounding box.
Download link: [UAV-ROD] (code: w0s1 )
The UAV-ROD consists of 1577 images and 30,090 instances of car category annotated by oriented bounding boxes. The number of images in training set and test set are 1150 and 427 respectively. We use drone to collect images and the flying height of the drone is between 30 meters and 80 meters. The image scenes include city roads, parking lots, residential areas, roadsides, etc. In UAV-ROD, The θ denotes the angle of car head direction and is increased in a clockwise manner. The w, h represent the width and height of the oriented box respectively. The average number of objects per image in our dataset is 19.08, of which the maximum number of objects per image is 104.
