Perceive Before Learning to Plan or Act - Multitask Perception Systems for Autonomous Agents

Attend a comprehensive seminar on advanced perception systems for autonomous agents in dynamic human environments. Explore the development of multitask models that integrate basic-level perception modules with high-level reasoning about human-to-human and human-to-scene interactions. Delve into cutting-edge research on end-to-end multi-object tracking, human motion forecasting, and human social grouping and activity prediction. Learn about the JRDB dataset and benchmark for training and evaluating end-to-end perception systems. Gain insights from speaker Hamid Rezatofighi, a lecturer at Monash University with extensive experience in computer vision, AI, machine learning, and robotics. Discover how these perception systems can enable safe navigation and interaction for embodied agents like mobile social robots in complex human environments.