Transition to the new object detection model version


Effective November 2020:

  • Common Data Service has been renamed to Microsoft Dataverse. Learn more
  • Some terminology in Microsoft Dataverse has been updated. For example, entity is now table and field is now column. Learn more

This article will be updated soon to reflect the latest terminology.

AI Builder is including improvements to object detection in a new model version. Your object detection preview models created before March 5, 2020, are deprecated. More information: Form processing and object detection preview models in AI Builder are deprecated

Starting March 5, 2020,you have to re-create your existing object detection models if you want to continue to use them. Models created after this date will automatically use the new model version. It's a good idea to transition to the new model version as soon as possible.

What do I need to do?

You must re-create object detection models created before March 5, 2020

  1. Identify any object detection models created before March 5, 2020, that you want to continue using.
  2. For each model, create a new object detection model with the same training images you used in the original model.
  3. After the new model is trained and published, make sure you update any apps or flows to use the new model.
  4. After you're done with your deprecated object detection model, you can delete it.

We understand that you might have spent significant time building your object detection models. Contact us if you need help with migrating larger complex models.

Why this change?

Object detection models are being upgraded for general availability. Your existing preview models won't be compatible with the new model version.

Until March 5, 2020, object detection models stored images as attachments in the Note entity. With this update, these models now use the new File and Image data types in Microsoft Dataverse, enabling a better and more optimized usage of capacity.

See also

Manage your model in AI Builder