Stream’s Analytics Platform

The key elements in the platform are:

  1. An analytics engine
  2. Account management
  3. The ability to create an Aapp
  4. The ability to use and existing Aapp
  5. An online store where Aapps are made available so others can use them
  6. Support for a broad range of hardware, through open source API


At the foundation of the platform is a world-class machine learning analytics engine. The analytics engine connects to data capture devices using our Cloud Connect software, available at no cost.

Stream has developed leading edge multi-band convolutional neural nets to automate the creation and ongoing execution of Aapps (Algorithm applications). These include specialized techniques to deal with both spatial and spectral data.


A free user account is used to create an Aapp, subscribe to an existing Aapp, or view an Aapps results (predictions).  All user accounts are accessed through a browser on a PC, tablet or mobile device.


Central to the platform is the ability to create an Algorithm application, or Aapp.  This is the algorithm or model which learns how to make predictions.   Data submitted from a relevant device is used to train the Aapp. The Aapp learns how to distinguish between images (or scans) which contain your target, and those which do not.

Aapps can be trained to solve complex problems, like predicting if a target is present or absent (classification), or how much of a target is present (regression). Examples of targets include disease in plants, nutrients in soil, nitrogen content in a leaf, or defects in manufactured goods.


Once created, an Aapp can be used privately or distributed publicly through Stream’s online store. Aapps made publicly  available allow anyone to subscribe to and use it – eliminating the need for them to create their own.  Creators of Aapps published for public use also receive a royalty based on subscriptions.

Anyone, anywhere, can subscribe to a public Aapp to predict answers to problems, or view test results provided by an Aapp.  For example, to determine nitrogen levels in a leaf, a picture of the leaf being tested is submitted to and then analyzed by the Aapp.  The prediction is made and results are displayed – all on a phone, tablet, or computer.


Once an Aapp has been created to solve a problem, it can be published so others can subscribe to it. Aapps are distributed and available for subscription in the online store. The price of each subscription is determined by the creator of the Aapp. Each Aapp offers a free trial tier to allow users to make a set number of predictions at no cost to help determine how well it will work for them.


To support the broadest range of data capture devices possible, an API is available to device manufacturers.  Many devices can find a new purpose as they capture data used to create an Aapp, or to capture data from samples which need to be analyzed. Using the ONSI, our open source API, makes this integration quick and easy.

Different problems are best solved with different types of devices. A picture is rich in data and provides some spatial reference. A spectrometer offers information across a wider spectrum but focuses on a smaller point. Appropriate devices include RGB cameras, multispectral and hyperspectral cameras, and spectrometers ranging from VIS to NIR. Here is more information on compatible hardware.