Why does the "import labels" feature in an Azure Machine Learning Labeling Project tell me that it finished successfully when it didn't import any labels?

Benedikt Schmitt 50

I have a labeling project that I started in CVAT which I want to continue in Azure Machine Learning. For this I uploaded the images to AzureML and created a data labeling project for them.

I exported the labels from CVAT as COCO-files and converted them into the jsonl-format so that I could upload them as an mltable. I have also used the same labels in my project that I used in CVAT.

Now I want to import these labels into my data labeling project so that I don't have to start completely from the beginning.

Every time I try to import the labels it tells me that the import was successful but that it also didn't import any labels.

santoshkc 12,185 Reputation points Microsoft Vendor

2024-04-12T11:48:18.4433333+00:00

Hi @Benedikt Schmitt,

Thanking you for reaching out to Microsoft Q&A forum!

If the import is successful but no labels are imported, it could be due to incompatible file formats or incorrect image paths.

To resolve this, ensure that the COCO files are correctly formatted and contain the labels you want to import, and that the image paths in the COCO files match the image paths in your data labeling project.

See: How to create image labelling projects

Also see: Troubleshoot problems during creation of a data labeling project.

I hope you understand! Thank you.
Benedikt Schmitt 50 Reputation points

2024-04-12T15:02:01.9266667+00:00

Thank you for the answer. I will try this solution on monday and come back to you.
Benedikt Schmitt 50 Reputation points

2025-02-05T10:59:54.6733333+00:00

@santoshkc

I have to revive this old question since the project is being continued now.

The answer above was most likely correct, that the image-paths were not correct. However is there official documentation on how the image-paths have to be constructed? I have tried several ways to give the paths but none of them worked.

It would be really helpful if Microsoft provided a template of how the data has to look in order for it to be successfully imported.

Currently I have a data labeling project with 3 images and 3 labels. The images are pulled from a Data Asset.
I also have labels for these three images in a .json-file with the coco-format. The image file-paths I have tried are:
-just the image names how they are called in the data asset
-relative path from the blobstore
-Storage URI-path for every image

After importing these labels I encountered two different errors:
-the import was successful however none of the images were labeled
-the import was successful but every time I want to check the labels it couldn't find the image and gave the following error:
"File not Found. Media could not be loaded. The file is missing from storage. It may have been deleted by the owner."
This is not due to access rights.

Share via

Why does the "import labels" feature in an Azure Machine Learning Labeling Project tell me that it finished successfully when it didn't import any labels?

Your answer