AWS DeepLens

 

• A deep learning-enabled camera for developers
• A wireless-enabled camera integrated with AWS Cloud
• Capable of delivering 100GFLOPS of computing power (1 billion operations per second)
• Contains sample projects at launch to get you started
• Optimized for Apache MXNet, TensorFlow, and Caffe
• Integrates with Amazon Rekognition for advanced image analysis

Common use cases

• Developing computer vision applications such as:
  • Face Detection
  • Activity Detection
  • Object Detection
  • Bird Classification 
  • Artistic Style Transfer

AWS DeepLens needs 3 AWS services to create a project:

• Amazon SageMaker 
  • Train/validate custom or pre-trained models 

• AWS Lambda
  • Preprocessing 
  • Capturing inference
  • Displaying output
• AWS IoT Greengrass
  • Deploys application project and Lambda runtime to AWS DeepLens 
  • Handles software and configuration updates

AWS DeepLens Device Library

• awscam module
• • • Runs inference code based on a project’s model.
• mo module
• • • Converts Caffe, Apache MXNet, or TensorFlow deep-learning model artifacts into AWS DeepLens model artifacts.
    • Provides optimizations for AWS DeepLens model artifacts.
• DeepLens_Kinesis_Video module
• • Can send video feeds from the AWS DeepLens device to Amazon Kinesis Video Streams.