As for the hardware, DeepLens is a 4 megapixel camera with 1080P video, 2D microphone array, Intel Atom processor and 8GB of memory for models and code.
The device runs Ubuntu 16.04, AWS Greengrass Core and optimized versions of MXNet and Intel clDNN libraries.
AWS outlined DeepLens at re:Invent and has received interest from educators, students and developers. For AWS, DeepLens is a way to put machine learning services in the field to find use cases. AWS DeepLens also ties into SageMaker and AWS Lambda. See: What is AI and machine learning?
Amazon added that it has added computer vision models using Gluon, SageMaker imports and tools to optimize models. DeepLens will also support Amazon's Kenesis video streams.
Dr. Matt Wood, general manager for machine learning services, said DeepLens makes machine learning less abstract and real world.
"This is a learning device. I can see developers inside the enterprise using machine learning in their own products," said Wood. "DeepLens allows developers to hone everyday skills and apply beyond in their organizations. With machine learning you need a lot of data. A video camera can capture things inside house, office and categorize them. There were a high number of devices have gone out to developers at re:Invent many with no machine learning experience."
Wood said developers would typically take a few different routes to get started in machine learning. Some developers would go the academic research route and learn novel methods, but these approaches would be hard to apply to everyday problems. Other developers would go with open source code and tinker, but have the same end challenge, said Wood, adding that DeepLens is aiming to bridge those gaps.
As for early projects, DeepLens has been applied to everything from book reading to games to categorizing skin lesions.