mittagessen

I pinned shapely to 1.7.x until we get a handle on things. 1.8/2.0 is a fairly large rewrite on their side so kicking the can down the road until things...

On 21/12/18 08:20AM, Stefan Weil wrote: > The ground truth data was produced at UB Tübingen with Transkribus. > The Transkribus PAGE export typically has rather bad line boxes, but...

Sorry, I forgot about it. Let's reopen and have a look. It's probably related to the arcane image loading routines that often returned different numerical ranges than a simple `np.array()`...

A separate validator is preferable as it would allow tools that use the API like escriptorium to validate a user-provided spec. In the `KrakenTrainer` object we'd have to instantiate the...

Are you segmenting the page? The exact command line you're running should throw an error (```Failed processing pagexml_trans.jpg: No line segmentation given. Add one with the input or run `segment`...

As ALTO doesn't define valid and invalid representations I'd prefer having something that is able to parse all 4 examples proposed in the ticket. AFAIK PageXML suffers from the same...

Hmm, I can't really reproduce it either on master or 4.2.0. The segmenter doesn't output any lines at all (as it should be) which is usually the reason that recognition/segmentation...

Thanks, this is great! The anti-aliasing was disabled because of B/W lines which obviously aren't anymore with it. Disabling it didn't seem to have any accuracy impact on grayscale models...

On 23/12/04 09:42AM, Robin Champenois wrote: > Indeed, that is something important, as CNN models can be very > sensitive to this kind of change. As the new `_rotate` uses...

On 23/12/05 01:16AM, Robin Champenois wrote: > Ok, thanks, I'll run it with BiblIA! (but it's a pretty big one, it > will take a bit of time) A subset...