Rendering bounding boxes produced by ML model results in application hanging

[KevinToepelt]

Describe the bug
Rendering an image with predicted bounding boxes causes the web application to load forever.

To Reproduce
Steps to reproduce the behavior:

1. Use a ML backend to produce a prediction, either through the actions menu, or through the labeling window.
2. Wait for the bounding boxes to appear
3. You will instead be met with a loading window, which does not resolve if the number of bounding boxes >= 1.
4. Reloading the application will result in the image (and bounding boxes) being rendered, but with a loading circle over them.

Screenshots
Screenshot of loading window shown below:

After reloading, I am met with this:

Environment (please complete the following information):

• OS: Ubuntu 24.04
• Label Studio Version 1.12.1

[sajarin]

Thanks for sharing this issue with the team @KevinToepelt. Does this happen with any image, regardless of size? Are there any logs that you can share from the docker containers?