The comparision to other works mentioned in paper

[pean1128]

Hi, I test the LayoutGMN and the results are so different. I find the related link of LayoutGMN has not released an official pre-trained model on UI dataset. Could you provide the pre-trained model used in your validation? Thanks a lot.

[lyf7115]

LayoutGMN estimates the final correspondence from the correspondence weight matrix (i.e., the attention weights).

Well, in our setting, after obtaining the weight matrix we use Hungarian algorithm to produce more reasonable correspondence results.

I have dropped an email including the pre-trained models to you.

[pean1128]

Thanks for your reply sincerely. 👍

[lyf7115]

https://drive.google.com/file/d/1-tLGVDn4I17fEY4QoQZdoF2H3ImIxANO/view?usp=sharing

[pean1128]

https://drive.google.com/file/d/1-tLGVDn4I17fEY4QoQZdoF2H3ImIxANO/view?usp=sharing

方便加一下微信吗，可以交流下~