Batch Processing about keras-ocr HOT 3 CLOSED

That's definitely part of the plan. You could do it directly with the detector (detector.detect allows you to pass multiple images to get text boxes for multiple images at the same time) and recognizer. That said, there are some known kinks to work out with the RNN layers in the recognizer that come up when conducting inference with larger batch sizes. Once that's fixed, I'd like to add a function to allow people to do inference across multiple images with batching in the pipeline.

My winter vacation is over so I can't contribute as frequently but watch this space!

EDIT: I would be remiss if I didn't add that PRs are welcome for this package and for this feature, in particular.

from keras-ocr.

This has been implemented as of 8de825e and is in v0.6.0. The README and examples have been updated accordingly. pipeline.recognize now takes lists of images instead of single images.

from keras-ocr.

I think I've addressed this request as discussed above. If you find that it does not meet your needs, please comment on this issue and we can revisit. Thank you!

from keras-ocr.