NAVER LABS Europe seminars are open to the public. This seminar is virtual and requires registration.
Date: 15th June 2021, 10:00 am (GMT +02.00)
Local-metrics for multi-object tracking
About the speaker : Jack Valmadre is a researcher and expert in object/person tracking at Google, France. He did his Ph.D. at Queensland University of Technology in collaboration with the Australian national science agency CSIRO, followed by a Postdoc at the University of Oxford. He regularly publishes at CVPR, ICCV, ECCV and other main computer vision venues. He has recently proposed new metrics for multi-object tracking and a study of state-of-the-art tracking methods. In general Jack is interested in the general principles of continuous and discrete optimization, theoretical computer science and programming languages.
Abstract: I will introduce temporally local metrics for Multi-Object Tracking. These metrics obtained by restricting existing metrics based on track matching to a finite temporal horizon, and provide new insight into the ability of trackers to maintain identity over time. Moreover, the horizon parameter offers a novel, meaningful mechanism by which to define the relative importance of detection and association, a common dilemma in applications where association is tolerable. It is shown that the historical Average Tracking Accuracy (ATA) metric exhibits superior sensitivity to association, enabling its proposed local variant, ALTA, to capture a wide range of characteristics. In particular, ALTA is better equipped to identify advances in association independent of detection.