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
553 stars 191 forks source link

Predicted output contains extra characters/missing character from tokens. #103

Open foramjivani opened 2 years ago

foramjivani commented 2 years ago

Thank you for the awesome solution. however, I am facing an error while predicting output as below : I have trained model for name and address key and I am getting validation accuracy for this as below :

+--------------------+----------+----------+----------+----------+ | name | mEP | mER | mEF | mEA | +====================+==========+==========+==========+==========+ | address | 0.920846 | 0.943102 | 0.931841 | 0.943102 | +--------------------+----------+----------+----------+----------+ | name | 0.965278 | 0.923588 | 0.943973 | 0.923588 | +--------------------+----------+----------+----------+----------+

but when I predict I am getting following output Rajesh Gangaram -- name

Yadav 374,shri ram chawk, omr road, chennai, tamil nadu, 600097 -- address

however Yadav is part of name.

Can you please guide me how to reduce there token misclassifications ?

Neelesh1121 commented 2 years ago

@foramjivani @wenwenyu same thing is happening with me also, please let me know if you have resolved this issue.