Amazon Web Services Inc. wants to give developers hands-on deep learning experience with its latest hardware. Announced at the AWS re:Invent conference in Las Vegas Wednesday, the AWS DeepLens is a wireless camera that developers can use to run deep learning and image recognition models.

The DeepLens camera is a collaboration between the Amazon.com Inc. cloud company and Intel Corp. that provides developers of all skill levels with the necessary tools to create artificial intelligence and machine learning products. It’s another example of AWS’s efforts to bring machine learning to the developer masses.

With the growth of machine learning in various consumer products, including smartphones and security cameras, the AWS DeepLens could help drive further innovation from developers. The DeepLens will retail for $249 and will ship on April 14, but is available for preorder now. It is currently only available in the U.S.

AWS DeepLens will work in a similar way as Google LLC’s Clips camera, an AI-powered hands-free camera, which the tech giant introduced in October. The Clips camera will also retail for $249, but a release date is still unknown. While Clips is marketed for consumers, DeepLens has been specifically targeted at developers.

“Now you have a platform where we can develop a special use case,” Naveen Rao, vice president and general manager of the Artificial Intelligence Products Group at Intel Corp., told SiliconANGLE. “Think of the camera as an extension of the cloud. It’s not a product, it’s a platform.”

AWS DeepLens hardware and software

The wireless programmable DeepLens device sports a 4-megapixel camera that can capture video at 1,080p HD resolution and audio with a 2D microphone array. The device is powered by an Intel Atom X5 processor and includes 8-gigabyte memory and 16GB of storage. The compact device also includes various ports, including microSD, Micro-HDMI and two USB ports.

On the software side, it runs Ubuntu 16.04 and is preloaded with AWS Greengrass, software that extends cloud capabilities to local devices. It can operate with open-source AI services such as Google’s TensorFlow and Facebook Inc.’s Caffe 2, as well as AWS’s new SageMaker, which was also introduced at the re:Invent conference.

Amazon SageMaker is a service that simplifies the deployment of AI models and provides the option to select a model prebuilt by AWS or developers can train their own models based on various criteria.

DeepLens pre-trained models

DeepLens can be programmed to do “almost anything,” according to AWS Chief Executive Andy Jassy, including automatically opening the garage door when the camera recognizes a specific license plate or sounding an alarm when the dog jumps on the couch.

Although developers can create their own deep learning models trained by Amazon SageMaker, they can also make use of the DeepLens pretrained model library. The model can then be deployed to the DeepLens device with a single click and developers can watch the results in the AWS Management Console in real time.

THE AWS pre-trained models offer a variety of image detection and recognition, including facial, cats and dogs and inanimate objects, like headphones and plants, as well as hot dogs (“classify your food as hot dog or not hot dog,” a popular internet meme). The pretrained models can also identify more than 30 activities, including brushing teeth and playing the guitar.

With reporting from Robert Hof

Images: AWS