Theelx

> There's also this fix for empty images that it took me a bit too long to discover. > > #11432 Oh thanks lol, I never noticed that! I had...

Also, I believe 1.4.11 could work too if ``strict=False`` was passed to ``albumentations.Compose``, as specified in the Core section of the [1.4.11 release notes](https://github.com/albumentations-team/albumentations/releases), though I haven't tried it yet.

Ok, I can confirm that passing ``strict=False`` to [mmsegmentation's codebase's initialization of albumentations.Compose](https://github.com/open-mmlab/mmsegmentation/blob/b040e147adfa027bbc071b624bedf0ae84dfc922/mmseg/datasets/transforms/transforms.py#L2361) results in it working with mmdeploy, so I'd think the same would happen with mmdetection.

If it helps, I was able to reproduce this when trying to sync files between an Ubuntu 22.04 computer and Linux Mint 22 computer after upgrading both to v1.28.0. It...

Hey, I also think this sounds super useful! I would definitely be happy to try and integrate this into jsonpickle itself, and it sounds like davvid would be too. I...

I'd be happy to take a look at the uncaught exceptions and be added to the bug tracker. However, I'll leave the final decision on the latter up to @davvid...

Ok, well in that case I can give you my email if davvid approves it, as keeping jsonpickle safe is more important than keeping my email safe :)

The test failures are from Ruff, from a previous commit having an unused variable. I'll make a PR to fix that now.

@dwagon If you could rebase this on the main branch which'll fix the tests, and if @davvid approves this, then I think this'll be ready to merge!

Oops sorry, I see now I forgot to review this. Either way though, it looks good to me, thanks for merging davvid and thanks for the PR dwagon!