Event-based vision sensors are based on biological principles and are therefore well suited to performing tasks similar to those performed by biology. The sensors provide sparse visual data with high temporal precision and low redundancy, which allows for efficient processing at low computational cost. The sensors themselves are low power and operate with low latency, making them a good choice for mobile battery operated platforms which must interact with the world in real-time. However, the data provided by these sensors differs significantly from conventional frame-based vision data and therefore requires new approaches to processing. In this talk I will describe some of the work in our lab on processing event-based vision data for tasks such as recognition, tracking, motion estimation, and pose estimation.

报告人简介：

         Garrick Orchard is a Senior Research Scientist at Temasek Laboratories and the Singapore Institute for Neurotechnology (SINAPSE) at the National University of Singapore. He holds a B.Sc. degree (with honours, 2006) in electrical engineering from the University of Cape Town, South Africa and M.S.E. (2009) and Ph.D. (2012) degrees in electrical and computer engineering from Johns Hopkins University, Baltimore, USA. He was named a Paul V. Renoff fellow in 2007, a Virginia and Edward M. Wysocki Sr. fellow in 2011, and a Temasek Research Fellow in 2015. He received the Johns Hopkins University Applied Physics Lab’s Hart Prize for Best Research and Development Project, and won the best live demonstration prize at the IEEE Biomedical Circuits and Systems conference 2012.  His research focuses on developing neuromorphic vision algorithms and systems for real-time sensing on mobile platforms. His other research interests include mixed-signal very large scale integration (VLSI) design, compressive sensing, spiking neural networks, visual perception, and legged locomotion.