Qualcomm and UvA: leading the way in deep vision
In 2015 the Qualcomm joined forces with the UvA to create a research center in the relatively new field of deep vision: the QUVA lab. Peter Rauber works at Qualcomm as a Senior Director of engineering, and is responsible for operation and development in the company's European research devision.
Using your smartphone to send an email, watching a movie on your tablet, getting a message on your smartwatch: big chance that, without realising, you are using technology from Qualcomm, the biggest chip provider in the wireless industry.
Deep vision for mobile devices
Deep vision is a combination of computer vision and deep learning. To teach a computer to see and understand the world around us is a complex and challenging task. Great advances are made with the help of new machine learning techniques and powerful computers. At the QUVA lab researchers want to find ways to make deep vision possible for mobile devices and the internet of things (IoT).
'A lot of software on your mobile device needs connection to the internet to work, a lot of the calculations are done in the cloud,’ says Rauber. 'But there are many reasons why this isn't always preferred, and you want the information to be processed at the edge: on your mobile device. Qualcomm is leading efforts in machine learning on the edge, a field we are very excited about. Running machine learning on device means better privacy, lower latency, reliability and efficient use of bandwidth. There’s a lot to explore in the field and applications in IoT, autonomous driving and speech recognition to name just a few.'
Working with world-renown researchers
Why choose the UvA for the collaboration? ‘In recent years Qualcomm acquired two UvA spinoffs, Euvision Technologies and Scyfer, so we already had a good relationship with the university. The three directors of QUVAlab, Cees Snoek, Arnold Smeulders and Max Welling, are very experienced and world renowned-researchers. They are also very inspiring people.'
Investment in talent
Three postdocs and nine PhD students work at QUVA lab and collaborate with the Qualcomm Research Netherlands team. Qualcomm Research Netherlands also offers internships where students can work on cutting-edge projects for four months. Besides that, QUVA hosts a Qualcomm-sponsored lecture series with recognised speakers in the field of deep learning.
'We hope to do something for the community, we also do a lot in STEM-education,’ says Rauber. 'Each year we offer fellowships for PhD-students of seven selected top universities in Europe. The three winners with the best research proposals receive $40,000 each and are assigned a Qualcomm mentor. The event is hosted at the UvA in May.'
Being located at Science Park, Qualcomm has a good chance to meet talented students at the Faculty of Science and consider them for permanent positions. 'Talent is scarce, and the demand is humongous, so one of our goals is to interest more students to pursue a career in machine learning.'
But according to Rauber the students also have something to gain: he describes the QUVA-collaboration as the best of both worlds. 'The researchers get to interact with engineers and scientists in an industrial research environment and learn to work within the constraints of practical solutions. We hope that this type of collaboration will contribute to more people pursuing research and a career in machine learning. With hundreds of millions of chips sold every year, Qualcomm engineers have a chance to make an enormous impact in the world.'