TGRSS ang DIOR remote sensing

DIOR is a large-scale benchmark dataset for optical remote sensing image target detection proposed by Han Jw's group at XIT. The dataset contains 23,463 images and 192,472 instances with an image size of 800×800 pixels and covers 20 object classes. Among them, the training set contains 5862 images, the validation set contains 5863 images, and the test set contains 11725 images. These 20 object classes are: Wind mill, Vehicle, Train station, Tennis court, Storage Tank, Ship, Harbor, Stadium, Ground Track Field, Golf course, Expressway toll station, Expressway service area, Dam, Chimney, Bridge, Overpass, Basketball Court, Baseball field, Airport, Airplane.