It’s been seven months since Amazon Web Services gave the first look at DeepLens — a $249 programmable video camera that runs machine learning models for object recognition — and now the device is finally here.
The camera, which starts shipping Thursday, is meant to put Amazon’s powerful machine learning capabilities in the hands of all developers. Powered by an Intel Atom X5 processor, DeepLens can recognize faces and objects, read text and more.
On the eve of its release, GeekWire got a chance to check out the device, first announced at the AWS re:Invent conference in Las Vegas last year. I made the trek down to Amazon HQ and entered a conference room in the Doppler building. The small white camera was already connected to a terminal, waiting for me to give it a shot. On a nearby table was an assortment of items — a toy boat, cat and dog stuffed animals, a wine bottle — that will come in to play later.
It’s a good looking little device. The square 4 megapixel camera sits atop a 4.5-inch tall box. On the back are ports for USB, microHDMI, speakers and a Micro SD card. It has 8 MB of memory to store models and code on the device. The device I had gets pretty hot after it’s been running for awhile, so be careful there.
It’s important to note that DeepLens is a developer-focused device. It does have some easy to use out-of-the-box applications for the everyday user, but folks with some technical skills who want to get their hands on Amazon’s machine learning capabilities will get the most enjoyment out of DeepLens. Additionally, viewing a stream of a project deployed to the device requires a micro HDMI cable and a USB connection to a keyboard and mouse.
My tour of DeepLens was led by Swami Sivasubramanian, vice president of Amazon Machine Learning and Jyothi Nookula, senior product manager for AWS. Despite covering the technology industry, my coding experience begins and ends with dropping pictures and links into my stories, so they had their work cut out for them.
Nookula says DeepLens is meant to be usable for developers of skill levels, including ones with zero experience with machine learning. Amazon boasts that users can open the box and deploy their first deep learning model within 10 minutes.
“As long as you can click some buttons, you’re good to go,” Nookula said. “That’s the reason we have these sample projects.”
Setup was smooth, and it was easy to deploy the sample projects on to the device. But when I took the device back to the office, I didn’t have the right cables to connect the device to my machine to stream the applications on my computer.
After running through the setup process, the first DeepLens project I played with was a pre-programmed object recognition program. When the camera turned on, it recognized me as a person and some of the toy objects for what they were. It had trouble with the wine bottle, instead focusing on a table behind it.
In addition to titles of objects, the device gives you a percentage of how certain it is that the object actually is what the DeepLens perceives it to be. The other out-of-the-box applications can recognize faces of people as well as dogs and cats, identify a variety of activities like playing a guitar, take a visual effect in a photo and super impose it on top of the video feed and, in a nod to HBO’s Silicon Valley, discern whether an item is a hot dog or not.
Though I only got a limited look at what the device can do, it’s easy to see the potential of this kind of technology. Retailers could use it to count guests throughout the day. One application can recognize unfamiliar faces that could be a sign of a break-in. The winner of a recent DeepLens developer competition took home the crown for building a project that can read children’s books.
Sivasubramanian, a 12-year veteran of Amazon, lives in the city of Sammamish, Wash. about 15 miles east of Seattle. The town is known for its bear sightings, so Sivasubramanian built a project that can recognize a bear and then sends him a text message alert of its presence.
Sivasubramanian said the camera is part of a mission to do with machine learning what AWS did for cloud computing: put powerful tools from one of the biggest tech companies in the world into the hands of developers of all skill levels. DeepLens offers developers another entry point into Amazon’s ecosystem, using many of the same tools deployed in other parts of the platform.
A camera is a pretty specific application of such a big idea. Sivasubramanian says Amazon picked this device from a list of several possibilities because the visual nature of it would be enticing to developers.
“Developers really love building solutions in a hands on fashion with devices,” he said. “Developers get even more excited when they can physically touch and program and build applications that they can track and show to their friends and family.
DeepLens is not Amazon’s first foray into cameras. The Echo Show and Echo Look both rely heavily on their cameras, and the Amazon Cloud Cam is part of a program to let delivery people drop packages off inside the home.
DeepLens is a pretty device, with some fun out-of-box experiences. But to unlock it’s full potential, you’ll need some technical skills, or at least a friend who will do everything for you.