LogisticsWhere: Zoom Instructor: Liangyan Gui TA: Shengcao Cao, Yunze Man Practice presentation: Zoom UpdatesOverviewSummary: Much of existing work in computer vision has focused on learning models of high accuracy. However, in real-world, resource-constrained scenarios, such as AR/VR, autonomous driving, and robots, being able to reduce latency and predict the environment?s dynamics is equally important. This advanced graduate course will cover foundation principles and recent progress of learning efficient and predictive models and their applications in domains such as vision, robotics, and NLP. We will investigate state-of-the-art approaches and a wide range of recent research topics, such as improving trade-off between accuracy and efficiency, predicting human motion/video/trajectory, anticipating action/event/intention, etc., in single- and multi-agent settings, and with multi-modal data. We will cover both the theoretical foundations and the techniques to build such practical systems, e.g., model compression, knowledge distillation, sequential modelling, predictive learning, multi-modal learning, etc. People
Academic Integrity Policy
|