Training Failed for Object Detection

florinbulg 1 Reputation point
2022-11-16T14:15:07.803+00:00

I'm trying to use Custom Vision for a quick MVP and I uploaded about 30 pictures. I annotated 3 tags each having 15 to 17 examples. I started with a "Logo" domain as this is the problem I'm trying to solve and at first it trained fine.
I then changed the domain to General (compact) in order to export the model and try to use it offline. The iteration failed. I did more attempts after this, including with the same configuration as in the first scenario (when it worked) but with no success. I keep getting "This iteration failed to train." message. I have to mention that I did not change the dataset at all. I am on the free pricing tier but far from reaching any limit.

What I tried:

  • quick training & advanced training with different domains
  • a new project with a similar dataset and only 1 tag

Is there a way to get more details regarding the error? I also discussed this on chat with customer support and none of their suggestions helped.

Azure AI Custom Vision
Azure AI Custom Vision
An Azure artificial intelligence service and end-to-end platform for applying computer vision to specific domains.
214 questions
0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Ramr-msft 17,611 Reputation points
    2022-11-17T12:28:30.897+00:00

    @florinbulg Thanks for the question. We can successfully train the object detection model. Please re-create the project with General (compact) in order to export the model. Compact domains are lightweight models that can be exported to iOS/Android and other platforms.

    https://learn.microsoft.com/en-us/azure/cognitive-services/custom-vision-service/select-domain