Open XYFFFFFF opened 3 years ago
Thanks for your interest. First of all, This result is NOT from my paper. I just implemented the research. You should read papers about attention, non-local blocks and other related ones.
Sorry for bother, I have another ques. During training, is it necessary to get a sketch, and a reference from the same original image?
This paper proposes the way. If we get sketch and reference images from different sources. There is no ground truth image.
When do the test, if the test image doesn' t neet to be shifted/elastic ?
It is depends on you. For an image, shifted and elastic is for only visualization. Check the code! Thanks
in your paper, you used three inear transformation matrices Wq, Wk & Wv, could u tell me how to calculate these matrices. Thanks a lot~