kraken

kraken copied to clipboard

This is an implementation for unsupervised pretraining of recognition model weights based on an image inpainting surrogate task that aims to reconstruct randomly sampled masked patches from the initial convolutional...

mittagessen

kraken already shows a good message for the segmentation model, but the corresponding message for the recognition model just says 'default': Loading ANN /home/stweil/src/github/mittagessen/kraken/kraken/blla.mlmodel ✓ Loading ANN default ✓ The...

stweil

Unable to segment specific images

27

This problem occurs with 11 out of a set of ~646 PNGs, all of which plopped out of the exact same processing pipeline, scanned on exactly the same hardware. Both...

sixtyfive

segtrain fails in v 4.1.2

6

Hi @mittagessen Just pinpointing a little oddity there : Running the following on 4.1.2 and torch 1.11: (Note that `ketos_sample` actually contains the `Italian` subdir from the [BiblIA dataset](https://zenodo.org/record/5167263)) ```python...

sven-nm

Hello @mittagessen, The training docs mention being able to slice off the final layers of a previously-trained model to speed up inference when alphabets are mismatched between the previous model...

christopherDT

Latest releases only produce an ABBYY file without any OCR results: ``` ```

stweil

masking of adjacent lines error

1

Since you mentioned some interest in this error, I’ll give full details here.  This is the result of a baseline segmentation using a custom model. The image file is...

jjarosch

The ocr results and reported model accuracy mismatch

10

I have trained a kraken recognition model on 20 PAGE XML files transcribed manually with Transkribus. Up to the evaluation stage everything seems to work just fine and `ketos test...

novacellus

[feature request] Training with Metal Performance Shaders

17

PyTorch recently introduced support for Apple's [Metal Performance Shaders (MPS)](https://developer.apple.com/documentation/metalperformanceshaders), see https://pytorch.org/blog/introducing-accelerated-pytorch-training-on-mac/. Using the new mps backend for PyTorch should accelerate training on Apple M1 machines a lot and is...

stweil

This pull request ist at the moment just to synchronize the discussion on how to proceed. It introduces a new CLI command, `ketos pagegen` that takes as input * a...

millawell