animegan2-pytorch icon indicating copy to clipboard operation
animegan2-pytorch copied to clipboard

Published 20 hours ago • verified publisher icon bryandlee

training code

Open Leocien opened this issue 3 years ago • 9 comments

Thank you for your stunning works, do you plan to release the training code (including the training code about the Face Portrait model)?

Leocien avatar Sep 07 '21 07:09 Leocien

Hi, I don't plan to release the code at this point (it's a total mess). I'm still working on improving the translation quality.

bryandlee avatar Sep 10 '21 10:09 bryandlee

I'm stoked to see it, not even just for using it with this project but just to figure out how to do this sort of task at all. Please tell us in this issue thread when you release it so I can get a notification. ♥

torridgristle avatar Sep 17 '21 13:09 torridgristle

Hi, I don't plan to release the code at this point (it's a total mess). I'm still working on improving the translation quality.

I tried to reproduce your Face portrait model by using the pix2pix method directly. My trainA is celebA dataset and trainB is the cartoon celebA generated by your Face portrait model, but my results are poor, would you like to share the training details to me? I also tried the pix2pixHD method, but the results are still not very good.

Thanks very much!

Leocien avatar Sep 23 '21 09:09 Leocien

Looks awesome! Have you been using vanilla AnimeGANv2 code for training? What dataset did you use for face2portrait model?

Sxela avatar Nov 07 '21 15:11 Sxela

Hi! Nice work, I'm also interested in the training code / datasets, specially for Face Portrait v2. Do you think it would be possible to revert the network and get realistic faces from painting / drawings?

Greg8978 avatar Nov 11 '21 00:11 Greg8978

Is there a way to train the network using pytorch?

tensorflow is a bit problematic

b4zz4 avatar Nov 15 '21 16:11 b4zz4

Hi, I don't plan to release the code at this point (it's a total mess). I'm still working on improving the translation quality.

I tried to reproduce your Face portrait model by using the pix2pix method directly. My trainA is celebA dataset and trainB is the cartoon celebA generated by your Face portrait model, but my results are poor, would you like to share the training details to me? I also tried the pix2pixHD method, but the results are still not very good.

Thanks very much!

Hello, Sorry to interrupt you suddenly, the cartoon celebA you mentioned refers to you used animegan's model to generate a dataset from celebA, or is the other cartoon datasets?

wzy0903 avatar Nov 29 '21 11:11 wzy0903

Hi, I don't plan to release the code at this point (it's a total mess). I'm still working on improving the translation quality.

Hi, Thanks for your impressive work.

Are there any updates or news regarding the training code?

ramtiin avatar Apr 14 '22 19:04 ramtiin

@bryandlee When will the training code be released. Or tell us about your ideas?

watertianyi avatar Mar 06 '23 11:03 watertianyi