1

I'm unsure how to associate the tagged regions with my original image file names.

Following the Custom Vision API documentation for GetTaggedImages request, I'm able to get a response with information about iteration id, created date, width, tagged regions, etc. However, there's no mention of the original image name.

If it's not possible to associate the original file name with the tagged regions, then I'll have to do my labeling offline (LabelImg for instance) and use the API to upload the tagged images - not the workflow I'm hoping for.

Thanks for any insights!

Aragon
  • 31
  • 6

1 Answers1

1

I received an answer to this question from a microsoft account representative. The image name must be associated with the "id" at the time of upload (using the custom vision api, not the web gui):

The service does not use filenames for tracking, it uses image IDs. When an image is uploaded, the ID is returned, if it's needed to associate file names to IDs that information should be captured on upload and stored externally in a file, database, etc.

Here's an example of how I got what I needed:

upload_result = trainer.create_images_from_files(project.id, images=tagged_images_with_regions)
for image in upload_result.images:
        print(image.image.id)
Aragon
  • 31
  • 6