Longtime readers of this column will know that I've been predicting the proliferation of visual intelligence in a wide range of products – including consumer electronics – for a few years now. Over that time, there have been a few high-profile successes of vision in consumer electronics. For example, the first-generation Microsoft Kinect, while not a hit with serious gamers, sold tens of million units and enabled many casual users to enjoy console gaming for the first time. There have also been some high-profile flops, including the second-generation Kinect and the Amazon Fire phone. Vision-based gesture control for consumer electronics – a hot topic a couple of years ago – has yet to capture consumers' enthusiasm.
To the casual observer, the lack of a significant number of successful vision-enabled consumer electronics products to date may suggest that vision technology is a passing gimmick – something like 3D TV – that was a hot topic last year but is now old news. My time at the Consumer Electronics Show earlier this month tells a different story.
What was most striking to me at CES this year was the sheer number of end products incorporating visual intelligence. In past years at CES I've seen many demonstrations of vision technologies – for example, small depth-sensor modules, eye-tracking software and gesture-control software. But for the most part what I saw in years past was technology demonstrations, not end products. This year, that changed – big time.
A drone that follows you as you snowboard, recording video along the way? Check. A head-up display for car drivers that "paints" navigation information directly on the road surface? Check. A smart surveillance camera that messages you when your kid gets home – or when an unknown person arrives? Check. This year at CES, for the first time, I saw dozens of end products incorporating visual intelligence. Of course, most of these products will fail. Most new products do. But I'm confident that a growing number of them will succeed.
My confidence stems from several considerations. First, just as eyesight is an invaluable capability for humans – enabling us to navigate through spaces, perform tremendous feats of dexterity, and communicate with one another – vision also adds fundamental, invaluable capabilities to many kinds of systems. As Stanford professor Fei Fei Li puts it very elegantly in this presentation, "Just as visual intelligence is important to nature, is a cornerstone of all intelligence."
This would be obvious to us except that, for decades, nearly all of our electronic devices and systems have lacked vision. While a few high-end systems like manufacturing inspection systems, surgical robots, and Mars rovers have done amazing things with vision, nearly all of the systems that we use every day – from our smart phones to our cars to our light fixtures – have been blind.
While it was the end products rather than the enabling technologies that stole the vision spotlight at this year's CES, there's no doubt that the enabling technologies are continuing to advance at an accelerating pace, thanks to increasing investment and competition. And these advances – in things like ultra-low-power processors that can perform face detection – will inspire and enable thousands of new vision-equipped end products.
Speaking of inspiring and enabling, if you're interested in learning about the latest technology, techniques, and applications for practical computer vision, you'll want to mark May 12th on your calendar. That's the date for the 5th Embedded Vision Summit – a conference for product creators who want to build better products using vision technology. Paralleling the growth in adoption of vision technology, the Summit has been growing rapidly, and this year's promises to be very exciting. Visit the Summit web pages in the coming weeks for details of the Summit technical program, which will be packed with world-class expert presenters. I hope to see you there!