The big picture: Amazon has many cloud services available and is making it easier than ever before to learn the basics of deep learning using the DeepLens camera. As a learning tool, DeepLens can teach you how to process video and images to obtain useful information.
Amazon is continuing to leverage the success of its web services and cloud resources. After originally announcing the AWS DeepLens camera at AWS re:Invent 2017, it is now available to US customers with worldwide shipments coming later on.
DeepLens shoots at 1080p with a 4MP sensor. The box attached to the camera itself is a single board computer with an Intel Atom Processor running Ubuntu 16.04. There is 8GB of RAM and 16GB of storage that can be expanded via microSD card.
On the back of the box, standard I/O includes audio out, two USB 2.0, a micro HDMI, and microSD card slot. A 2D microphone array is also available for voice and audio analysis. Dual-band Wi-Fi allows for streaming to and from cloud services.
Although DeepLens is not really an impressive camera in terms of imaging capabilities, it can process video and extract data in real time. Integration with Amazon Rekognition allows for complex image analysis.
A large number of Amazon web services are supported by DeepLens. AWS IoT, Amazon SQS, Amazon S3, DynamoDB, and more all seamlessly integrate. One of the goals for this new camera is to allow "developers of all skill levels to get started with deep learning in less than 10 minutes."
For hobbyists with a little more experience or professional developers, customer deep learning models can be created through Amazon SageMaker. Out of the box, the camera can detect around 20 different objects. Utilizing a face detection module, approximately 30 different actions
AWS DeepLens is available now with a suggested retail price of $249.99.