ocrd_anybaseocr icon indicating copy to clipboard operation
ocrd_anybaseocr copied to clipboard

block segmentation: non-text classes and prebuilt models

Open bertsky opened this issue 4 years ago • 0 comments

In e941321a507ce9f4f6d6416117e441124605748a it seems 3 non-text classes arrived: ImageRegion, TableRegion and GraphicsRegion. However, the Config.NUM_CLASSES remained the same, and equally the provided block_segmentation_weights.h5 still have only 1+14 classes:

>>> import h5py
>>> f = h5py.File('block_segmentation_weights.h5', 'a')
>>> f['mrcnn_class_logits/mrcnn_class_logits/kernel:0']
<HDF5 dataset "kernel:0": shape (1024, 15), type "<f4">

So @khurramHashmi what am I to make of this? Did you forget to publish/upload your new model file?

I would really like to see images and tables detected here.

bertsky avatar Feb 03 '21 10:02 bertsky