AWS wants to make every developer a machine learning expert with SageMaker and DeepLens camera

At AWS re:Invent, CEO Andy Jassy announced a new machine learning service called SageMaker, along with a deep learning-enabled camera.

Video: The biggest challenges and opportunities facing Amazon Web Services

SageMaker, an end-to-end machine learning service from Amazon Web Services (AWS), will make it easier for everyman developers to build, train, and deploy machine learning models.

The service was announced by AWS CEO Andy Jassy at the 2017 AWS re:Invent conference in Las Vegas. It starts with an authoring component that handles data cleaning and processing, according to a blog post.

It also offers scalable model hosting and training as well, the post said. The goal is to simplify the process for working with machine learning and democratize deep learning models.

SEE: Cloud computing policy template (Tech Pro Research)

In addition to built-in algorithms and one-click training, SageMaker has a feature called Hyper Parameter Optimization, Jassy said. This allows developers to check a box at the beginning of tuning their model and it will find the best parameters for your machine learning model.

Along with SageMaker, Jassy unveiled DeepLens, a wireless HD video camera that has on-board compute optimized for deep learning. DeepLens also has AWS GreenGrass on it so it can run Lambda triggers. It integrates with SageMaker and has tutorials, demos, and pre-built models included.

DeepLens has built-in video recognition tools and it is essentially a piece of hardware that can be used to help developers hone their skills in deep learning. It is available for preorder for $249.

Continuing the video trend, Jassy announced Amazon Rekognition Video, which provides real-time batch video analytics. It can detect objects in a video, detect inappropriate content, and even detect celebrities. The service also automatically timestamps everything it identifies to make the data easier to work with.

To further the usefulness of the timestamped data, Jassy detailed Amazon Kinesis Video streams to allow for secure, serverless ingestion and storage of video, audio, and other time-encoded data.

For audio, a new service called Amazon Transcribe will provide automatic speech recognition to help convert speech to text. This could make audio data more searchable and easier to use in other applications. It uses machine learning to add punctuation and formatting. It can recognize multiple speakers, supports telephony audio, and will support custom vocabulary. Transcribe will be available in English and Spanish to start.

To translate text between languages, Jassy showed off Amazon Translate. This service will offer real-time translation, batch analysis, and automatic language recognition, Jassy said.

For natural language processing (NLP), Amazon Comprehend will provide a fully-managed NLP service. It detects entities, phrases, language, and sentiment. Jassy said that it can search through millions of documents to determine key topics.

The 3 big takeaways for TechRepublic readers

  1. At re:Invent, AWS CEO Andy Jassy unveiled SageMaker, an end-to-end machine learning service that could democratize the technology, along with a deep learning-enabled camera called DeepLens.
  2. Other new services like Amazon Rekognition Video and Amazon Kinesis Video will use machine learning to help developers get more out of time-stamped audio and video data.
  3. Jassy also announced new NLP services, a real-time translation service, and Amazon Transcribe, that can transcribe audio into text.

Also see


Amazon's Matt Woods demos the DeepLens camera at re:Invent.

Image: Conner Forrest/TechRepublic