By Brenda Shen
When hearing the phrase artificial intelligence (AI), many immediately tend to think of the inevitable autonomous robot rise expected in the next few decades. While robots and smart machines make up a sizeable portion of the AI sector, AI encompasses much more than just the dreaded robot invasion. McGill’s Centre for Intelligent Machines is a prime example of the wide array of research taking place in machine and technology development, showing the versatility of artificial intelligence applications.
Formed in 1985, McGill’s Centre for Intelligent Machines (CIM), centered in McConnell Engineering Building, is McGill’s contribution towards promoting research on intelligent systems. Encouraging inter-departmental and inter-faculty collaboration, the CIM hosts fifteen different labs that are comprised of computer scientists, software engineers, and mechanical engineers.
Dr. Martin D. Levine obtained his B. Eng. and M. Eng. degrees from McGill in 1960 before obtaining his Ph.D in London. He is now a professor in McGill’s Electrical, Computer, and Software Engineering department where he is also researching computer vision, image processing, and AI. Several of his ongoing projects relate to visual surveillance. His focus is on monitoring the behavior of people and he is currently developing technology to better aid in the detection, recognition, and characterization of the objects he is tracking. For the visual surveillance of people, Dr. Levine focuses on the detection and recognition of the face and the body. This application has the potential to be incredibly important, as video surveillance systems are widely used and can provide valuable information to law enforcement and security forces.
A member at the Levine Lab, PhD student Mehrsan Javan Roshtkhari, is developing an algorithm in real-time anomaly detection, which is the identification of unusual objects and suspicious behaviors, that could be incorporated into an automated surveillance system. The computer science behind this detection is incredibly complex, encoding the video surveillance feed into compactm spatio-temporal volume arrangements where a probabilistic framework is used to calculate specific regions.
Another student at the Levine Lab, Mahannad Elhamod, also works in the detection of suspicious behaviors by working to annotate real-time visual surveillance through the categorization of their actions. While confirming these algorithms are successful and accurate in their detection of odd behaviors, there is a key challenge of AI development: ensuring that these programs continue to work regardless of difficulties in video quality. Issues such as shadows, occlusions, lighting changes, and non-static backgrounds all provide significant hurdles for these researchers and troubleshooting such problems take up a sizeable portion of time spent on AI projects.
Many other exciting initiatives are taking place at the CIM. The Shared Reality Lab (SRL), led by Dr. Jeremy R. Cooperstock, focuses on achieving the virtual reproduction of real data with as little distortion and change as possible. This can encompass data types including audio, video, force-based technologies, and human signals. The SRL has many innovative projects that span a variety of fields including fitness, medicine, and communication. Other research endeavors at CIM include the Robotic Mechanical Systems Laboratory and the Mobile Robotics Lab, with research topics aligning more with machine learning and the feared AI takeover. The CIM is a great example of the exciting progress in AI research contributed by McGill. With the recent news of Dr. Joelle Pineau, a professor in the School of Computer Science, being appointed as the head of Facebook’s new Montreal AI Lab, McGill continues to display its talented team of AI researchers making impacts in the field.