ADT: Object Detection Challenges
Detect, recognize and track 3D objects from Project Aria glasses
Current Challenges Overview
The ability to detect objects, and predict where those objects lie in 3D space (aka ‘spatialization’), is crucial to many applications of AR and AI. For mixed reality devices, these capabilities could allow users to identify objects in their environment, surface relevant information about those objects, and even visualize changes to those objects in realtime.
The 2023 Aria Digital Twin (ADT) challenges aim to accelerate research into these core capabilities, serving as a catalyst for the broader research community.
ADT Challenge One: 3D Object detection & tracking with scene generalization
Given the training data consisting of a list of moving objects and recorded sequences ground-truthed in two distinct scenes, your task is to generate accurate 6 Degree of Freedom (6DoF) poses for the same set of objects within recordings from a completely new and unseen scene.Participate in the challenge on Eval.ai
ADT Challenge Two: Few-shots 3D Object detection & tracking
Provided with a short period of ground-truth data for a previously unseen object within each sequence, your task is to leverage this training data to accurately detect the same object and predict its 6 Degree of Freedom (6DoF) poses for the remaining duration of the sequence.Participate in the challenge on Eval.ai
Awards for the challenge
First place: $10,000 and a certificate. Sponsor may also choose to publish the technical report of the first place submission.
Second and Third places: certificates and recognition on Challenge leaderboard. Sponsor may also choose to publish technical reports of second and third place submissions. Additionally, teams with the best performing or noteworthy submissions may be invited to present their work at CVPR in June 2024.
Update: The challenge deadline has been extended to 23:59 (Pacific Daylight Time) on Friday March 22, 2024. The new ceremony for challenge winners will be CVPR 2024.