Orientation detection

[micycle1]

Add an orientation module, which measures image orientation, to identify incorrectly orientated images in particular. We'd then have the option to rotate Lap previews of affected images, or rotate and correct the underlying image file.

Promising ONNX option here: https://github.com/duartebarbosadev/deep-image-orientation-detection

Others:
https://huggingface.co/Chuckame/deep-image-orientation-angle-detection
https://github.com/parsapoorsh/resnet-ixion

This is cool too, but more on object orientation within space: https://orient-anythingv2.github.io/

[julyx10]

Really appreciate the suggestion and the research on model options!

Thought about this one — I don't think it's worth building in right now. Most people using Lap are shooting on real cameras, and those almost always write correct EXIF orientation, so this mainly helps with scans/screenshots/downloads, which is a pretty small slice of most libraries. Adding an 80MB model + detection pipeline for that felt like overkill.

If this turns out to be more common than I think, happy to revisit — probably as an opt-in tool rather than something that runs automatically on import. If it does happen, duartebarbosadev/deep-image-orientation-detection looks like the right pick (98.82% accuracy, MIT license, already shipping in a similar app).

Leaving this open though 👍

[micycle1]

Yes I was envisaging it as opt-in (run manually under the module's view).