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

UnionLayer好像写的有问题? #75

Closed ChenCongGit closed 3 years ago

ChenCongGit commented 3 years ago

N个文本行的x和x_gcn不需要完全对齐吗?这里似乎只对x进行展开去padded的部分,但是x_gcn还是直接reshape(B,NT,-1),这样x_gcn的节点完全没法和x的文本行对齐了,而且后面x_gcn = x_gcn.reshape(B, N T, -1)[:, :max_doc_seq_len, :]不太对啊。

wenwenyu commented 3 years ago

Hi, 谢谢提醒,此部分已修复。