AWS Launches DeepLens AI Camera for U.S. Developers

The Amazon Web Services (AWS), DeepLens camera device with built-in machine learning capabilities is now available for sale.
Each DeepLens device costs $249. According to Jeff Barr’s blog post, shipping is currently only available to U.S.-based customers. However, “additional destinations” are being considered.
DeepLens was first announced at the AWS reInvent conference in November. It allows developers to take photos of real-world objects, and then easily incorporate them into machine learning projects.
AWS DeepLens claims that it allows developers of all skill levels to start deep learning in under 10 minutes. It provides sample projects with hands-on examples and can be started running with just one click.
DeepLens comes pre-built with deep learning models, but developers can also import their models from Amazon SageMaker’s machine learning platform. AWS AI executive Matt Wood demonstrated how SageMaker can help to create a machine learning model that recommends music based on past song selections. This was the first DeepLens demo. Wood then used DeepLens to feed the SageMaker model into DeepLens. He simply took photos of his face next the album covers to create “reviews”. DeepLens was able to identify Wood’s feelings for each album by his expression and assigned reviews accordingly.
DeepLens can also be integrated with other AWS services such as AI solutions such as Polly, Rekognition, and the Lambda programming environment.
The DeepLens camera. Source: AWS. The DeepLens camera itself is small, measuring approximately 140 millimeters high, 47 millimeters wide, and 94 millimeters in depth. The 4MP camera can record 1080p video. It supports Wi-Fi connectivity, a 2-D microphone array, and two USB 2.0 ports.
DeepLens inside is powered by an Intel Atom processor and runs Ubuntu OS-16.04 LTS. It has 8GB RAM and 16GB expandable memory.
Barr claims that AWS has added many new features to DeepLens over the preview phase. These include support for the Caffe, TensorFlow deep learning frameworks and support for more MXNet layers. Integration with Amazon Kinesis Video Streams is also possible. A new sample project can detect head positions.