Different performance on similar images

[Nell_Klinger]

I have three images from the same (printed) source, two of which Leo transcribed wonderfully, and one where it says it has transcribed it, but the transcription is blank. The pages are in a similar format / resolution, so I’m confused as to why it sometimes worked and sometimes didn’t.

[Jon]

Hey Nell! There is a known issue with the current model where it intermittently ignores printed text as Jack explains here:

Transcription failures

Thanks both for sharing, this is really useful feedback! I’ll add a little to what Jon says. For your images Amelia: I’ve had a look through the logs and the transcription jobs were technically successful (i.e. no software bugs), but the AI failed to return text in a large proportion of your images. I see that your images are mostly or entirely printed text – I believe what’s going on here is that Leo has been optimised to extract handwriting from images and, when presented with printed text, a…

We’ve spent the last couple of months cleaning up the data and are currently training a new version of the model which we think will no longer have this pathology. So long as all goes well it should be released this week