Overview

Custom Vision Service brings the power of machine learning to your apps

Custom Vision Service is a tool for building custom image classifiers. It makes it easy and fast to build, deploy, and improve an image classifier. We provide a REST API and a web interface to upload your images and train.

What can Custom Vision Service do well?

Custom Vision Service is a tool for building custom image classifiers, and for making them better over time. For example, if you want a tool that could identify images of "Daisies", "Daffodils", and "Dahlias", you could train a classifier to do that. You do so by providing Custom Vision Service with images for each tag you want to recognize.

Custom Vision Service works best when the item you are trying to classify is prominent in your image. Custom Vision Service does "image classification" but not yet "object detection." This means that Custom Vision Service identifies whether an image is of a particular object, but not where that object is within the image.

Very few images are required to create a classifier -- 30 images per class is enough to start your prototype. The methods Custom Vision Service uses are robust to differences, which allows you to start prototyping with so little data. However, this means Custom Vision Service is not well suited to scenarios where you want to detect very subtle differences (for example, minor cracks or dents in quality assurance scenarios.)

Custom Vision Service is designed to make it easy to start building your classifier, and to help you improve the quality of your classifier over time.

Release Notes

Dec 19, 2017

  • Export to Android (TensorFlow) added, in addition to previously released export to iOS (CoreML.) This allows export of a trained compact model to be run offline in an application.
  • Added Retail and Landmark "compact" domains to enable model export for these domains.
  • Released version 1.2 Training API and 1.1 Prediction API. Updated APIs support model export, new Prediction operation that does not save images to "Predictions," and introduced batch operations to the Training API.
  • UX tweaks, including the ability to see which domain was used to train an iteration.
  • Updated C# SDK and sample.

Known issues

  • 1/3/2018: The new "Retail - compact" domain model export to iOS (CoreML) generates a faulty model which will not run and generates a validation error. The cloud service and Android export should work. A fix is on the way.

Next steps

Build a Classifier