In the information era, new technologies with world-changing potential are being invented every day. Although implementing them into practical use is no easy task, and it may take years or even decades to do so, the relentless development of technology is making everything possible.
Like cutting-edge fields such as 3D printing and augmented reality, computer vision, which sounded too futuristic to be practical when it emerged in the 1980s, is coming out of the research labs and going mainstream as both the technology and the market mature. As a technology with enormous application potential, computer vision is attracting major companies, like Microsoft, Yahoo, Google, Facebook, as well as a growing number of start-ups.
Shanghai-based startup Yitu Technology is one of the new industry entrants. Yitu operates a cloud-based visual recognition engine that enables computers to detect and recognize faces and cars. The system was first applied to security surveillance to help authorities identify persons of interest in criminal investigations and to track traffic violations.
Leo Zhu, who gained a post-doctoral fellowship on computer vision at MIT, founded the company with high school friend Chenxi Lin, a former cloud computing technology director at Alibaba. Yitu has come a long way since its establishment in 2012. Leo explained that basic face recognition systems can search faces from galleries of millions, while Yitu’s engine can find one face from hundreds of million with higher accuracy in real time.
Positioned as a research- and technology-driven company, Leo invited his Ph.D adviser, Professor Alan Yuille, to give a mini course on computer vision in Shanghai last week. Professor. Professor Yuille was part of the MIT artificial intelligence research arm and has made significant contributions to the emergence of modern computer vision. He is well known for his work on image edge coding, object recognition, and his long term interest in unifying biological and computer vision modeling and technology.
Professor Yuille started out with a BA in mathematics at the University of Cambridge and then was supervised by Professor Stephen Hawking while undertaking a PhD in theoretical physics. His current role is professor in the department of statistics at UCLA and he also holds concurrent appointments in the departments of psychiatry, psychology and computer science.
Yuille believes that the ultimate goal of computer vision is to create intelligent systems which can understand the world as well as or better than human beings. These systems will help us to understand our brain as well as the world we live in. It is estimated that 40-50% of neurons in the cortex are involved in vision, so deeper investigation in this area will help us better understand the brain.
The human brain can handle multiple visual missions in the blink of eye, such as finding the object in a picture while describing its the contents and details at the same time. However, machines only can manage limited missions. Yuille says the problem for researchers is how to use existing systems to process enormous amounts of data to a level similar to human brains.
He considers the fragmented and dispersed field of computer vision one of the factors restricting it from reaching its full potential, in part due to its interdisciplinary nature and rapid growth. There is vast duplication of effort, and not enough building on the work of others. Companies should focus on different verticals rather than duplicating each other’s work.
Computer vision is a rapidly developing technology with enormous applications, and the gap between academic research and industry application is narrowing, especially in the last five years, he noted. Major breakthroughs have been achieved by leading tech companies based on the latest academic research. “Maybe technology will become mature enough to spare the backing theoretical studies someday, but that’s not our concern for now, because there are still lots of deep and fundamental researches to do.”
Yuille himself sits at the junction of research and industrial application, as he is also a co-founder of Blindsight, a startup which has been acquired by Amazon. The company was founded as a part-time job when he was a research fellow and had substantial free time. He added that it was fun to do something “real” after years of academic research. When talking about characteristics that he valued most in entrepreneurs, Alan said they were the ability and confidence to carry things through and the ability to see what the future is going to be.
Professor Stephen Hawking recently warned that intelligent machines could pose a threat to humanity. “The development of full artificial intelligence could spell the end of the human race”, he said in a BBC interview. When being asked to comment on the idea, Yuille said this fear may concern that the emergence of intelligent machines and robots will lead us to a jobless future. “In one of his speeches thirty years ago, Hawking said machines may replace physicians after we solved problems in theoretical physics. I think he has had this concern decades ago.”
Editing by Mike Cormack (@bucketoftongues)