Model Builder

Quick, easy, no experience required.

Sign Up

Our platform makes it easy to perform supervised deep learning.  We offer binary classification, multi-class classification, and regression models.  Binary classification is often used where a yes/no or true/false answer is required, while multi-class classification deals with identification of multiple things, such as species identification of plants or the identification of automobile brands.  Regression can predict things like value or percentage.

Supported Data Types

At this time we support regular RGB images, multi-spectral images, and spectrometer scans in our automated machine learning model builder.  This is because we have deep expertise in reflected light and how to process this type of data.

If you have another type of data you would like to use to create machine learning models, contact us and we’ll work with you to integrate it into our model generation process.

When capturing data, our platform has the added benefit of our Cloud Connect software.  Cloud Connect enables us to control spectrometers, multi-spectral cameras, and regular imaging cameras like the Canon Powershot SX70HS.

You can also directly upload any compatible images to our web interface for use as training information.

If you are using hardware that is not yet supported, please contact us, we would be happy to discuss supporting new hardware.

Working With Training Data

To get training data into the platform, we allow direct upload, API upload, and using hardware integrated with our Cloud Connect software.

The training data used for each training run is recorded so you can see what changed between training runs that produced better or worse results.

Data Tagging

Once you have got your training data into the system, it will need to be tagged so the machine can learn about the various attributes you’re interested in.  Simply set up a variety of classes or values and assign them to the appropriate training data.


Through the machine learning platform, training is a one button push. All hyper-tuning is completed in the background and the best model is trained and tested, with all of the results displayed quickly and easily when they are available.