Face Dataset Questions

[leoliu37]

However, if you're in a rush, I believe split the dataset randomly and run re-train our model produces a similar result reported in the paper.

Thanks, I have obtained the CelebA database, but I wander if you can provide the corresponding relationship between each image of your dataset matrix in"face.h5" and the attribute filename in CelebA. Additionly, as you discribed in paper <Detach and adapt>, half of photos are converted into sketch domain by pix2pix. But how do you train the photo-sketch model?

Originally posted by @leoliu37 in https://github.com/Alexander-H-Liu/UFDN/issues/3#issuecomment-435775328

[bkompa]

Also interested in the correspondence between the celebA dataset and the face.h5 matrix. @Alexander-H-Liu could you provide either this correspondence or the mapping between elements of the different domains? So that we can know the ground truths

[WQYuan]

So amazing work! I'm also interested in how to make face.h5 file.Please support more details if you can.Thanks!