Import Tropy selections as Omeka image annotations

[allanaaa]

I've tested how a Tropy item with rectangular selections on the image imports into Omeka.

The coordinates in Tropy don't seem to be quite correct - you can see I drew rectangles around the printed photographs but I got rectangles full of text instead. (And in the wrong orientation.) So, that'a bug.

But, I mean to say that the new Image Annotate module (https://omeka.org/s/modules/ImageAnnotate/) makes similar rectangular selections of a particular image, along with textual notes attached to each rectangle. So, instead of exporting separate image files, perhaps you could just send the coordinates to match the module's data format.

You can take a look at the manual to see how things work: https://omeka.org/s/docs/user-manual/modules/imageannotate/

Again, I'm not sure if we're accepting this information via API yet, but we probably could, if you thought this was worth offering.

[inukshuk]

Sounds good! If the annotations can indeed be created via the API then we can definitely do this.

Can you share the original photo in the screenshot above with us? Did you rotate in Tropy before or after making the selections? Judging by the dimensions of the selections, it clearly looks like the orientation of the internal GPU texture is different than expected. It could be caused by the EXIF orientation or other metadata embedded in file.

[allanaaa]

It seems to be happening pretty consistently across different images. I'll upload a few here - some I've rotated within Tropy, some are unrotated. I haven't done any edits outside of Tropy.

This last one, that seemed to work correctly, was rotated using the button in the top bar.
Ones that didn't work are the ones I rotated by right-clicking on the image tile in the left bar and selecting "rotate left" or "rotate right", or using the command-bracket shortcut.

[inukshuk]

Thank you! There were actually two separate bugs, both in combination with EXIF orientation. Tropy displays the image correctly, according to how the camera was held when taking the photo, but when saving the thumbnail or extracting the image, the coordinates of the selection were applied to the original image. Depending on the orientation, the extracted region was in the wrong place.

[allanaaa]

Thanks for that! Let me know if you want me to test anything.

As per the exporting of annotations, maybe you can assess whether there would be call for that from your users? I can close this and open a separate issue if this is getting cluttered.