![]() ![]() Our customers bring problems which will give Applied Scientists like you endless opportunities to see your research have a positive and immediate impact in the world. AWS is the world-leading provider of cloud services, has fostered the creation and growth of countless new businesses, and is a positive force for good. As part of the team, we expect that you will develop innovative solutions to hard problems, and publish your findings at peer reviewed conferences and workshops. Our team of scientists have developed the algorithms and models that power AWS computer vision services such as Amazon Rekognition and Amazon Textract. ![]() The AWS AI Labs team has a world-leading team of researchers and academics, and we are looking for world-class colleagues to join us and make the AI revolution happen. September 1: Challenge submission deadline.August 15 17: Camera-ready-papers deadline.July 20: Posting of private leaderboard.July 15 August 2: Final submissions deadline and Paper submission deadline.We invite you to challenge the baseline and submit your results and papers by July 15. ![]() The workshop’s organizers envision this workshop as the first in a series of annual workshops, which will both catalyze and publicize research on autonomous flight - ultimately leading to even safer skies for drones. For this challenge, there is a budget of false identifications for each frame. The goal is to maximize the ratio between the number of the detected airborne objects, on a per-frame basis, and all the airborne objects that should be detected. The other challenge is frame-level object detection. The tracker must also have a false-alarm rate no higher than one false alarm every two hours. One is encounter-level detection, which means successfully tracking an airborne object for three seconds - before the object is within 300 meters of the drone. Participants in the challenge may choose to solve either or both of two challenges. In total, AOT includes close to 164 hours of flight data: 4,943 flight sequences of around 120 seconds each, collected at 10 Hz in diverse conditions. The images have been manually annotated to indicate the locations of the airborne objects.Īirborne objects usually appear quite small at the distances that are relevant for early detection: 0.01% of the image size on average, down to a few pixels in area. In addition to the so-called planned aircraft, AOT contains other unplanned airborne objects. Their trajectories were designed to create a wide distribution of distances, closing velocities, and approach angles. To generate the flight sequences in the AOT data set, two aircraft were equipped with sensors and flew planned encounters. By sharing our dataset, defining the problem and evaluation criteria, and hosting a challenge, we hope to expose the computer vision community to a relatively fresh area of autonomous flight applications, while emphasizing the real-world requirements for safe autonomous flight. Against this backdrop, the safe operation of drones requires fully autonomous and robust sense-and-avoid systems. In recent years, applications of drones have grown to include infrastructure inspection, emergency response support, agricultural and environmental surveys, and package delivery, among others. The workshop, which will be held virtually on October 11, 2021, will also feature talks by invited speakers, including Amazon researchers (Amir Navot) and academics (Laura Leal-Taixé from TU Munich, Pascal Fua from EPFL, and Andreas Geiger from the Tübingen Institute). Amazon Prime Air has also sponsored $50,000 in prizes for challenge participants, including $15,000 to the first-place finisher and a $2,500 prize for the “most creative” safety solution, as determined by the judges. ![]() The top six finishers in the challenge will give short talks during the workshop. The challenge, which launched in April, is to track airborne objects across successive frames of video, with the ultimate goal of obstacle avoidance. The yellow bounding boxes around airborne objects are the ground truth for learning and testing. Sample data from the Airborne Object Tracking dataset. ![]()
0 Comments
Leave a Reply. |