search

wenwenyu / PICK-pytorch

Code for the paper "PICK: Processing Key Information Extraction from Documents using Improved Graph Learning-Convolutional Networks" (ICPR 2020)
https://arxiv.org/abs/2004.07464
MIT License
560 stars 193 forks source link

Bug about normalizing the rel_feat #97

Open tankgit opened 3 years ago

tankgit commented 3 years ago

data_utils/documents.py: Line 297

lack of a pair of parentheses, I think it should be

feat[:, :, i] = (feat[:, :, i] - min_value) / (max_value - min_value)

Suppose the result in the paper was based on the code with the bug, I would expect the better performance of the approach after fixing it.