ValueError: bbox must be normalized

[robert-go]

Thanks for your awesome library
I'm trying to recolor iris
But some image have error: ValueError: bbox must be normalized
Here is the google collab to reproduce: https://colab.research.google.com/drive/1UL7F1f-8fyKXBoUg2YcVjAFDN_dfIcIS?usp=sharing

[patlevin]

Thank you for reporting this issue.

The library was a bit too strict with its coordinate system. It didn't anticipate images where not all of the returned face landmarks are actually located within the image (e.g. zoomed-in faces).

This will be fixed in the next release.

[patlevin]

The new version with the bugfix is available on PyPi.
Your Google collab example works without errors now after fixing the import error here to:

from fdlite.examples.iris_recoloring import recolor_iris

Many thanks for providing an example for proper repro 👍

[robert-go]

@patlevin thank you for a quick update
It's worked
Can you update the tutorial to fix import recolor_iris ?
https://github.com/patlevin/face-detection-tflite/blob/main/docs/tutorial.md#fun-with-iris-detection

[patlevin]

Can you update the tutorial to fix import recolor_iris ?
https://github.com/patlevin/face-detection-tflite/blob/main/docs/tutorial.md#fun-with-iris-detection

The updated tutorial was part of the latest release - check the link.

[robert-go]

I mean the import recolor_iris line.
It hasn't been updated yet

[patlevin]

Nice catch! 😄 It's fixed now.

[robert-go]

Thank you @patlevin
We can close this issue