SKYSCENES Leverages New Algorithms to Improve Safety for Autonomous Flying Vehicles
Ph.D. student Sahil Khose worked with Assistant Professor Judy Hoffman to curate SKYSCENES, a new benchmark dataset that provides well-annotated aerial images of cities that computer vision algorithms can use to operate autonomous flying vehicles. Photos by Kevin Beasley/College of Computing.
An artificial intelligence (AI) training dataset developed at Georgia Tech is setting a new standard for the safety and reliability of autonomous drones and flying vehicles.
SKYSCENES compiles more than 33,000 annotated computer-generated aerial images. With applications in urban planning, disaster response, and autonomous navigation, the dataset trains computer vision models to better detect and identify objects in aerial images, which can be challenging for existing AI models.
Read the full story to learn how School of Interactive Computing Ph.D. student Sahil Khose and Assistant Professor Judy Hoffman developed this groundbreaking dataset to pave the way for the future of autonomous aviation.