[cs-talks] Ajjen Joshi Oral Exam TODAY 12 - 1 in CAS 226
Harrington, Jacob Walter
jwharrin at bu.edu
Fri May 5 10:08:00 EDT 2017
Human Motion Analysis in the Age of Depth-sensing Cameras
Ajjen Joshi, Department of Computer Science, Image and Video Computing
TODAY May 5th, 12 – 1pm in CAS 226
Abstract: The launch of consumer depth-sensing technologies combined with efficient human pose estimation algorithms has enabled a proliferation of datasets comprising human gestures, actions and activities, along with novel machine learning algorithms that facilitate their analysis. One of computer vision's biggest success stories in the past decade has been the development and deployment of the Microsoft Kinect. In this talk, I will first present the efficient human pose recognition algorithm that lies at the heart of the Kinect and illustrate the impact it has had in the computer vision community, emphasizing research involving human motion. In the second half of the talk, I will remark on the coupling of large scale human skeletal datasets collected using the Kinect with modern deep-learning techniques, focusing particularly on the work by Neverova et al., who developed a multi-modal deep model for the task of gesture segmentation and recognition, along with recent work by Liu et al. who proposed an LSTM model that leveraged both temporal and spatial dynamics of the human skeleton in motion.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks