Fausto Morales

I'm not sure what's going on here -- but you could start debugging by taking a peek at what's going on inside `get_maximum_uniform_contour` by doing something like the following. Note...

The length of the predicted sequence is always equal to `input_width/(pool_size**2) - rnn_steps_to_discard`. You are getting a model weights shape error because the Kurapan weights use a Spatial Transformer Network...

To increase detection speed, just a couple ideas: - Use a GPU (😂 ) but even that is still not going to get to Tesseract speed, I don't think. -...

Something is going wrong with the box warping process. Please provide a sample image to reproduce the issue. My apologies for the inconvenience.

Interesting! Reading that error, it seems that ONNX doesn't support automatic padding while also using dilation. You can avoid using automatic padding by setting a fixed input size so that...

Great question -- weakly supervised training has been very challenging to implement (even more than I anticipated). The original authors have not provided an official implementation (not that they should...

In case it's useful, I gave an example on how to adapt the existing binary label-oriented metrics for a multi-class setting in https://github.com/tensorflow/tensorflow/issues/37256#issuecomment-751130406. HTH!

I've published a Keras port of the model that uses the original PyTorch weights (they get converted to Keras) as part of a package for building an end-to-end OCR pipeline....

Great idea! I don't really have the bandwidth to work on this right now but a PR would be welcome!

This was a thing in v0.1, but the shortcut was something like Ctrl+Shift+I or something and I felt like that was crazy. Any suggestions for a natural keyboard shortcut for...