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

Refine: auto fetching images and annotation files instead of specifying. #44

Closed tengerye closed 3 years ago

tengerye commented 3 years ago

Auto parse the extension and fetch the data, assuming the extensions are consistent within a dataset (note train_dataset and val_dataset are two datasets).

@dbobrenko @wenwenyu Hi, if possible, would you please check if it fits what are required? Thanks.

dbobrenko commented 3 years ago

So it parses the extension of the first image and then continues to use it? Looks good to me, maybe later we can even try to make it work with images with different extensions in the same folder

tengerye commented 3 years ago

@dbobrenko Yes, you are right.

For different extensions, we can try to search for each image every time. But it does not seem elegant for me and I am wondering how the case may happen.

wenwenyu commented 3 years ago

@tengerye That's a good modification and I will merge it to the current version.

@dbobrenko To make it work with images with different extensions in the same folder is good advice. Bu as @tengerye says, it probably is a time-consuming operation. Maybe later we can find a better solution.