maam m research scholar in csir-csio research lab this all video of your playlist is really very helpful for me in my work..i love your teaching style and specially u have implement code with minor details, architecture and concept that great thing..m extremely thankful 4 u keep it up..
Hi, I don't think tranposed convolutions and nearest neighbour upsampling are same thing, in earlier we have a learned kernel where the later one is parameter-less.
Mam you explain this tough concept in very easy and effective way. I read the paper 2 Times and gave up. But now I understand the idea fully. Thanks alot. Want to join your institute as a student.
See Data Augmentation simply means generating new samples on the basis of your current samples which are present in your dataset. Generator is a part of GAN network and with the help of generator you can generate new samples but there is no sense to use generator as a replacement of data augmentation techniques because generator will take so much time to get trained to generate fake data which looks similar to your real data. Its better to use generator for other useful tasks which I have explained in 1 of my videos on GANs. And yes you can use data augmentation to increase the data in GANs also
It's a really helpful video to expalin this network, thank you. And I've got two questions,1. how to update the MLP(mapping network) during training? I understand that we can update generator easily by gredient descent, but how to update the MLP and what's the loss function for it? 2. Is the input of MLP Z is a random vector or an image? Thanks again for this nice video.
I don't think we need to update weights of mapping net our random z might have few dominant features which may lead to entanglement..to avoid this, we pass it through mapping net that gives w which is independent of input data distribution and hence, a disentangled latent And Z is random vector here...but still some recent gan papers use latents of some images Both may work...but style gan authors experimented with random z afaik
one more important thing mam, ur teaching thumbnail/vdo on the tutorial is bit larger so it hides the info behind...well its good to see u closer but not good for missing any detail of tutorial afterall u do it very nicely so plz cut it bit smaller, thanks
the slides are being cut by your video many times. Pls place ur video such that slides are seen fully. If u feel doing so reduces the slides size, show it in slideshow mode instead.
Hi, Sorry for the inconvenience and Yes, I am aware about this problem. At that time, I was new in making videos, So this happened but if you will check the recent videos there this problem is solved.
I don't currently share the PowerPoint slides, but I'm glad you found the video helpful. If you have any questions or need further assistance, please feel free to ask, and I'll do my best to help.
@@CodeWithAarohi Yes maam this whole playlist is one of the best and easy to understandable I have got a mini Project of single Image deraining i.e removing rain streaks from an image using GANs. need your help in this
Hey, apprecheate the video idea, but i felt a bit bestranged when you repeat 5 times what everybody can see in the image already. actually what would be super disirable at that moment is to see how for example a "style" looks like at this point, when its fed to AdaIN... stuff like that..
Sometimes, when you are trying to understand something new, your brain kind of lags while processing the information. For me, this repetitiveness focuses attention on this new thing and gives some time to think.
I just love this video and how you clearly explained all the concepts by constantly making reference to the traditional architecture!
Thank you very much
Glad you enjoyed it!
You explained very well and comprehensibly, thank you.
Glad it was helpful!
Thank you very much! Your explanation was excellent! It is helping me a lot with my graduation work
Glad it helped!
maam m research scholar in csir-csio research lab this all video of your playlist is really very helpful for me in my work..i love your teaching style and specially u have implement code with minor details, architecture and concept that great thing..m extremely thankful 4 u keep it up..
Glad my videos are helpful 😊
😍😍😍😍😍😍😍😍
very well explained mam
Thanks a lot 😊
This video is very helpful and you explain the network very clearly, thank you!
Glad it was helpful!
I wonder why the research paper wasn't as explanatory as this video is. Thank you very much for sharing this video, it was 29 minutes well spent.
Glad it was helpful!
What is work of adain
Hi, I don't think tranposed convolutions and nearest neighbour upsampling are same thing, in earlier we have a learned kernel where the later one is parameter-less.
Mam you explain this tough concept in very easy and effective way. I read the paper 2 Times and gave up. But now I understand the idea fully. Thanks alot. Want to join your institute as a student.
Gladmy video helped you 😊
great explanation, thank you very much !
Glad it was helpful!
Great explanation!
Glad it was helpful!
good explain🎉
Thanks for liking
nice explanation, thank you madam. i'm waiting for your explanation and implementation video of diffusion model
As soon as possible
@@CodeWithAarohi ❤️❤️
Mam in which GAN we creates many similar images from a single image
Very helpful and informative, thank you!
Glad it was helpful!
Hi i didnt the implementation of StyleGAN. I got only StyleGAN ADA can u post the link for StyleGAN
I haven’t done that yet. I directly did StyleGAN2ADA
This video is very good, thank you!
Glad you liked it!
Face recognition using style is possible mam?
What is Style Mixing in style gan?
Hi Aarohi, in generator generates fake image ,is this same as augmentation ? can we use augmentation if we have less data in GAN?
See Data Augmentation simply means generating new samples on the basis of your current samples which are present in your dataset. Generator is a part of GAN network and with the help of generator you can generate new samples but there is no sense to use generator as a replacement of data augmentation techniques because generator will take so much time to get trained to generate fake data which looks similar to your real data. Its better to use generator for other useful tasks which I have explained in 1 of my videos on GANs. And yes you can use data augmentation to increase the data in GANs also
thank you so much for this video, can you change the size of latent space (z) or not ?
Can I get this slides?
It's a really helpful video to expalin this network, thank you. And I've got two questions,1. how to update the MLP(mapping network) during training? I understand that we can update generator easily by gredient descent, but how to update the MLP and what's the loss function for it? 2. Is the input of MLP Z is a random vector or an image? Thanks again for this nice video.
I don't think we need to update weights of mapping net
our random z might have few dominant features which may lead to entanglement..to avoid this, we pass it through mapping net that gives w which is independent of input data distribution and hence, a disentangled latent
And Z is random vector here...but still some recent gan papers use latents of some images
Both may work...but style gan authors experimented with random z afaik
one more important thing mam, ur teaching thumbnail/vdo on the tutorial is bit larger so it hides the info behind...well its good to see u closer but not good for missing any detail of tutorial afterall u do it very nicely so plz cut it bit smaller, thanks
Noted ! Thank you
Thank you mam I was waiting for your videos.your giving excellent content .please make videos for all types of GAN mam
Glad to hear that and sure I will make videos on all types of GANs soon. Keep Watching
Can u provide the ppt of stylegan
Cool
Thanks!
thanks
Welcome
LOVE VIDEO !
Thank you
Thank you!
welcome
looking good in this new hairstyle
Thank you
the slides are being cut by your video many times. Pls place ur video such that slides are seen fully. If u feel doing so reduces the slides size, show it in slideshow mode instead.
Hi, Sorry for the inconvenience and Yes, I am aware about this problem. At that time, I was new in making videos, So this happened but if you will check the recent videos there this problem is solved.
Can you share the ppt Maam Please
I don't currently share the PowerPoint slides, but I'm glad you found the video helpful. If you have any questions or need further assistance, please feel free to ask, and I'll do my best to help.
@@CodeWithAarohi Yes maam this whole playlist is one of the best and easy to understandable
I have got a mini Project of single Image deraining i.e removing rain streaks from an image using GANs. need your help in this
please send pdf of styleGAN
Hey, apprecheate the video idea, but i felt a bit bestranged when you repeat 5 times what everybody can see in the image already. actually what would be super disirable at that moment is to see how for example a "style" looks like at this point, when its fed to AdaIN... stuff like that..
Thankyou for appreciation and sharing your thought. Also Sorry for inconvenience. Will take this in mind while making my next videos :)
@@CodeWithAarohiI personally find repetition key to learning new concepts
Sometimes, when you are trying to understand something new, your brain kind of lags while processing the information. For me, this repetitiveness focuses attention on this new thing and gives some time to think.
The important content is not showing due to your PICTURE flash on your video.
Sorry for the inconvenience. I took care of this issue in my latest videos.