Computer Vision News - August 2018

Who is not playing their cards as they should? Everyone is playing, in one way or another. The hype is so much that everybody has to put out some sort of press release or marketing collateral that they are playing. I would say that after 2017, almost everybody has jumped in the market in one way or another. There are some who are late. There are some who are behind. If you look at the hardware landscape, Qualcomm and AMD are somewhat behind. If you look at the application companies, the automotive companies have done very, very well in terms of incorporating computer vision. The security industry, because of the fragmentation and different nature of the applications, has been behind. But I’m pretty sure that a lot of companies have started to invest. What about software? Who is surprising you, at the moment, for lagging behind? The software problem is twofold. One is that there are applications that need software on the device. Let’s call that edge computing. Then there are problems that can be done on the cloud. We can call that cloud computing. I think industry, in general, was a little bit late in terms of jumping into the edge computing market and figuring out how best to monetize it. The edge computing software market is quite fragmented. There are a large number of applications with widely different needs so I don’t blame companies for not investing heavily. With NVIDIA, their lowest power shipping product - the Jetson - is about 7.5 to 10 watts, and it is still not suited for many of the edge applications. Intel has acquired Movidius, so they can cater to sub 5W device category consisting of drones, robots and other applications. Both these companies essentially give out software stack for free to developers so they don’t monetize. On the cloud side, Amazon was the first to jump in. They have created a lot of APIs. There’s a lot of infrastructure on the cloud. Google came in after that, and then IBM. And then Microsoft. I think now they are all pretty much level in terms of where they’re at… A lot of startups. Literally, there are thousands of startups. In fact, NVIDIA’s inception program, I would say, is the best tracker of the world industry. They have some 3,000 to 4,000 companies in their database. It’s a very large list of companies trying to do different and interesting things in computer vision. As far as software is concerned, today, all the cloud companies are pretty close to each other. Edge companies... that is a different theater. There is no dominant software player, per se. There are a lot of small players in that area, but nobody has been able to figure out how to create one solution for middleware and monetize it. Let’s play a game. You are a young software engineer going for your first job. You have an offer from each one of the top companies. Which one should you go work for? [ laughs ] Good question! It depends on my goals as an engineer, right? If I want to learn a lot of things and get exposed to the really good technology out there, I would go to Facebook, Google, NVIDIA or any of these big 29 Guest Computer Vision News Guest