MRM & MAC - MARCS Research Meeting and MARCS Afternoon Colloquium 2nd March 2021

Join us at the next MARCS Meeting as we host an eventful MRM and MAC program commencing with talks from Dr Bharath Ramesh, Titus Jayarathna, and invited guest Associate Professor Arindum Basu

NOTE: Both sessions will take place via Zoom ID 839 6390 0819. Password marcs

PROGRAM

11am-12pm - MARCS Research Meeting

Topic: Mid-level Representations for Neuromorphic Vision Sensors
Speaker: Dr Bharath Ramesh

Many successful computer vision models transform low-level features such as Gabor filter responses into richer representations of intermediate or mid-level complexity. This process can often be broken down into two steps:  (1) a coding step that transforms the features into a representation better suited to the task, and (2) a pooling step that encapsulates the coded features over larger receptive fields. Mid-level representations will be especially useful to the visually sparse and often disjoint information recorded by event cameras. The applications of mid-level feature extraction include tracking fast moving objects using event-based cameras by making use of these locally consistent intermediate representations.

Topic: Optimising ADF military working dog performance through next-generation monitoring systems

Speaker: Titus Jayarathna

Military working dogs (MWDs) are essential members of the Australian Defence Force (ADF). The ADF needs a practical, automated, robust system that integrates with existing equipment to detect early indicators of impaired performance in MWD, enabling handlers to make decisions about work and rest periods and alerting them to signs of injury and impending danger of exhaustion or heat injury. Body temperature alone is unlikely to be an adequate indicator of the risk of imminent heat injury in MWDs. Sensor arrays that monitor individual MWD behavioural and physiological states are required to enable a ‘smart’ system sensitive to the current physiology of each dog. WSU in partnership with UNSW and USyd currently developing a system to monitor physiological signs of MWDs with novel sensor arrays, and integrated data analysis techniques to provide handlers with a system to facilitate rapid, informed decision-making to optimize MWD performance. Beyond MWDs, early warning systems for impaired dog performance or health would be valuable for many applications. Working dogs are relied upon in the police, correctional facilities, service dog sector, stock management, and racing. For some sports, where the welfare of dogs may come into question, such as sled racing and greyhound races on very hot days, it could become mandatory for participants to use such a system to ensure the health and welfare of the dogs involved.

1pm-2pm – MARCS Afternoon Colloquium

Topic: Neuromorphic 2.0: Brain-inspired AI for Edge Computing
Speaker:  Associate Professor Arindum Basu (Nanyang Technological University)

In this talk, Associate Professor Arindum Basu will introduce the concept of “Neuromorphic” Engineering (NE) that aims to develop neuroinspired circuits and systems for efficient intelligence in edge computing. Next, I will show two concrete examples of NE research done in our group. The first deals with developing intelligent cortical implants for brain-machine interfaces by integrating a low-power machine learner in the implant for decoding subject’s intention—this opens up the option of getting the next 10X increase in the number of accessible neurons. In the second example, I will demonstrate the opportunity of creating a scalable Internet of Video Things (IoVT) by using a neuromorphic camera as the front-end sensor. By being sensitive to only changes in temporal contrast like the human retina, such imager systems allow ultra-efficient sensing and processing compared to conventional frame-based cameras. Finally, I will close the talk with some examples of future directions.

The zoom ID is: 839 6390 0819

Link to zoom https://uws.zoom.us/j/83963900819?pwd=bDl5cVVkQWV2d1g4R2xaRFBtYTJiQT09