FAAM-satellite moving obejct detection dataset

Content:This study utilizes two satellite video datasets derived from Jilin-1 imagery (ground sampling distance: 0.92m, frame rate: 10 fps):VISO Training Set:70 video clips (512\u00d7512 pixels\/frame) totaling 27,434 frames.720,152 vehicle instances annotated with axis-aligned bounding boxes.Test Set :2,255 frames (1024\u00d71024 pixels\/frame) with 155,987 instances, featuring improved annotations over the original VISO test split.Both datasets cover several square kilometers per frame and focus on moving ground vehicles, supporting research in large-scale object detection and tracking in satellite videos.