Custom Vision Training Failed Error

Nathan Sokalski 4,121

I am trying to train a Custom Vision project for the first time (I have trained other projects before, so I know how to train). However, I continue to receive the following message:

"This iteration failed to train."

This is happening for Quick Training & Advanced Training with budgets ranging from 1 to 5. What is the problem? Thanks.

Nathan Sokalski 4,121 Reputation points

2020-07-18T22:47:52.2+00:00

I am also going to mention that I am using the Standard Pricing Tier (Pricing Tier: S0), and that all my training images have <200 regions, and I have >15 images for each tag. I have 13 different tags. As best I can tell, I meet all the requirements from the following page:
https://learn.microsoft.com/en-us/azure/cognitive-services/custom-vision-service/limits-and-quotas
Is there any way to determine WHY the iteration failed? Thanks.
R S 0 Reputation points

2024-01-25T08:35:57.5+00:00

Can't train just yet Your project can't be trained just yet. Make sure you have at least 15 images for every tag.Images were added as per lab...
R S 0 Reputation points

2024-01-25T08:37:20.3766667+00:00

Can't train just yet Your project can't be trained just yet. Make sure you have at least 15 images for every tag.This is the error .

1 answer

Ramr-msft 17,616 Reputation points

2020-07-20T16:42:09.217+00:00

@Nathan Sokalski Thanks for the details. Can you please check the Boxes’ relative width and height wrt image width and height matter more than absolute number of pixels.Generally speaking, it is better to have boxes not too big or too small. Can you please add more details about the location that you are using and if possible please share the samples to check.
Please sign in to rate this answer.
Nathan Sokalski 4,121 Reputation points

2020-07-23T23:49:03.283+00:00

The tag boxes in my training images are somewhat small, but I am not sure how to avoid that in my scenario. The project is for recognizing the dots on dominoes, and if I made the boxes much larger they would include multiple & varying numbers of dots (therefore making the data not useful). I could, supposedly, make the images larger, but that would mean training with images significantly different from the ones that will be submitted for testing by the app, which would probably make the resulting data less useful. What should I do?
Sign in to comment