AIML Research Seminar: Computational Algorithms for Human Behaviour Analysis - From Research Endeavor to Industry Relevance
Minh provided an overview of his research endeavours and interests, aiming to spark discussions about potential collaborative ventures.
He then delved delve into the challenges associated with transitioning research algorithms to industry, particularly those related to cost and accuracy. He shared firsthand experiences in developing cutting-edge surveillance technologies that tackle these issues, and discussed the design and implementation of a product for real-time monitoring of residential buildings utilising 9,000 video streams, featuring AI models that efficiently support up to 150 camera streams per AI accelerator, leading to a 30% reduction in hardware costs. He also introduced an innovative method for self-supervised model improvement, which enables AI systems to learn autonomously and adapt continuously to specific scenes over time, thereby enhancing both accuracy and performance.