Thanks so much man for explaining all of this. I was starting to read some papers on NST and didn't really know which papers to read until I came across this in-depth video. Keep up the good work!
Btw, if you found it useful it would really mean a lot if you share it so that others can see it! It's really tough with TH-cam preferring huge channels over the small ones (which does make sense on their side I totally agree, but this one is will get there! They just don't know it).
You mention GANs at the end and I love what styleGAN2 can do, however memory requirements for training GAN models restrict me from tinkering with them. Really good stuff here, been a year since I dove into this stuff and your explanations are awesome.
Hello dude！ Could u tell me how to train a set of style pictures (e.g. lots of pic from an artist), many thanks, training on single pic yield perfect result.
I guess I forgot to mention that one, thanks for letting me clarify, NPR stands for non-photorealistic rendering and it's a branch of computer graphics. NST is a technique inside this NPR field. That's how all those concepts tie together. 😁
Thanks! Not as much I'm much more into computer vision but I do have pretty strong background in mathematics/ classic ML/deep learning so I guess changing application area goes a bit smoother.
Thanks for inspiring and guiding with this indepth knowledge!
Amazing VIdeo! Best one in the series.
Thanks so much man for explaining all of this. I was starting to read some papers on NST and didn't really know which papers to read until I came across this in-depth video. Keep up the good work!
🙏
Great video! Very useful and detailed review of NST field, keep it up!
Incredible series!
Awesome really glad you found it useful! It's still a work in progress. Either next or the follow-up video will continue with the series!
Btw, if you found it useful it would really mean a lot if you share it so that others can see it! It's really tough with TH-cam preferring huge channels over the small ones (which does make sense on their side I totally agree, but this one is will get there! They just don't know it).
Thanks for the video! Keep up the good work!
Thanks a bunch for the support, will do!
You mention GANs at the end and I love what styleGAN2 can do, however memory requirements for training GAN models restrict me from tinkering with them. Really good stuff here, been a year since I dove into this stuff and your explanations are awesome.
Thanks a bunch, glad you find it useful! I'll create a whole series on GANs. It will probably be even longer than NST.
very good video. thank u so much
You're welcome! Thanks for the feedback
Hello dude！ Could u tell me how to train a set of style pictures (e.g. lots of pic from an artist), many thanks, training on single pic yield perfect result.
20:30 what does NPR stand for?
I guess I forgot to mention that one, thanks for letting me clarify, NPR stands for non-photorealistic rendering and it's a branch of computer graphics. NST is a technique inside this NPR field. That's how all those concepts tie together. 😁
A+
is there a colab with your implementation?
I've got 3 NST repos check them out on my GitHub:
The original one is here: github.com/gordicaleksa/pytorch-neural-style-transfer
Awesome video! keep it up 💪
Thanks a lot! Glad you like them!
Amazing job, best neural style transfer channel, are you also are proeficient with nlp?
Thanks! Not as much I'm much more into computer vision but I do have pretty strong background in mathematics/ classic ML/deep learning so I guess changing application area goes a bit smoother.