Curriculum Computer Science and Innovation for Societal Challenges, XXXVII series  Grant sponsor Università degli Studi di Padova Supervisor s Lamberto Ballan Contact sourav.das@phd.unipd.it

Project description

Prediction of human trajectory and forecasting human pose dynamics jointly is a fundamental requirement for several applications in robotics, autonomous driving, surveillance systems, etc. Though the problem sounds intriguing, it is extremely challenging in real-world scenarios owing to the different factors involved. Humans are intuitively social agents, and they are able to effortlessly conceive a detailed level of semantics from the scene, which contributes to making swift decisions for their next movements. To accurately forecast their pose dynamics, one primary factor is the social interactions in the scene and the influences the human joints have on each other. Similarly to precisely predict the future trajectory of a pedestrian, semantic and social information from the scene are extremely necessary. My Ph.D. project is to push the boundary of the current state of existing solutions for human pose and trajectory forecasting one step forward, especially towards more practical scenarios in-the-wild. The current advances in computational intelligence and learning, in addition to the increasing availability of road imaging and 3D human posing datasets and simulators, are both motivating this research project.