ACFR Seminar Series: An Adaptive Approach to Human-Machine Collaborative 3D Perception

When: Thursday, 27 August, 1:00pm AEST

Where: This seminar will be presented at the ACFR seminar area, J04, Level 2 (Rose St Building) and partially online via Zoom. RSVP

Title: An Adaptive Approach to Human-Machine Collaborative 3D Perception

Speaker: Ze’Ev Krischer

Abstract:
3D Perception models perform well in-distribution but tend to be fragile and degrade in new domains, collecting enough labelled data to fix this problem can be expensive. Many of these domains have human experts that could be used to fill the gap, but most systems either ignore them or their input as implicitly correct by default. Instead, our system realistically treats the human as a collaborator that is capable of making mistakes and considers how the AI system should adapt to achieve the best combined performance. In this talk I’ll walk through the system and demonstrate how devastating these systems on simulated users alone tends to lead to assumptions about behaviour that is not representative of real human users. I’ll also briefly touch on our ongoing study and what our future direction from it will be.

Bio:
Ze’ev is a PhD candidate in Human Robot Interaction and 3D Perception at the Australian Centre for Robotics supervised by Dr Mitch Bryson, as well as CSIRO Robotics’s HRI Team supervised by Dr Hashini Senaratne. He received a Bachelors in Mechatronic Engineering at the University of Sydney in 2023.