Action Recognition for Behaviour Understanding from Video in 2020
Automated Behavior Understanding is presently an actively researched area of Computer Vision. It is of interest for applications such as autonomous vehicles, health care, HCI, video summarization, etc. We may break down the behavior understanding task into several levels to make it simpler. At the first level, basic motions, poses, gaze may be recognized. At the next level, we may consider activity recognition approaches. Finally, we may further assign associations for group dynamics, predict crowd motion, interpret scene contexts or even intent from visual cues, essentially unfolding a higher degree of semantics along with temporal information. This talk will explore the latest developments and challenges in this area of research, with a focus on action recognition.
Tanmana Sadhu is a Computer Vision Engineer at Huawei Canada's Vancouver Research Center where she conducts research and development in video understanding. Tanmana graduated from University of Victoria, located in beautiful British Columbia, Canada, with a Masters in Electrical Engineering in 2016. She has since then worked in diverse domains of Machine Learning and Data Science spanning ocean research, banking and smart devices at various organizations.