Hi @Huage001 I read the paper and found that mean variance norm mean 'mean-variance channel-wise norm' works quite like instance norm. Can you explain to me why use mean-variance-norm function instead of instance norm? Thank you so much.

Hi , thanks for your code. As for the code in inference_frame.py, the line 110 in class AttnAdaINCos G_norm = torch.sqrt((G ** 2).sum(1).view(b, 1, -1)) b, _, h, w = F.size() F = F.view(b, -1, w * h) F_norm = torch.sqrt((F ** 2).sum(1).view(b, -1, 1)) F = F.permute(0, 2, 1) S = torch.relu(torch.bmm(F, G) / (F_norm + 1e-5) / (G_norm + 1e-5) + 1) # S: b, n_c, n_s S = S / (S.sum(dim=-1, keepdim=True) + 1e-5)

1. i think the F_norm and the G_norm should be calculated by sum(2) not sum(1)
2. why you add the 1e-5 in three places in code?

Hi , thanks for your code. I have two questions about it.

1. I don't understand what self.device means here, it doesn't seem to be defined.

2. In addition, I would like to ask whether local feature loss is taking up too much memory.Because I ran out of memory when I tried to import this module into other code.

Thanks!

In my screen, the lambda_content is always zero, i find the lambda_content is zero by default, i dont know how to set it , could you tell me how to set? Thanks!

Thanks for the great work! It looks like this repo currently only hosts code and model for image style transfer. Could you provide a pre-trained model (and perhaps test scripts) for video style transfer as discussed in the paper? Thank you.

The results of your paper is interesting.

So, I want to train your model from scratch.

Before training, I want to know approximate training time when you trained your model.

Thanks.

Hey @Huage001! 👋

This pull request makes it possible to run your model inside a Docker environment, which makes it easier for other people to run it. We're using an open source tool called Cog to make this process easier.

This also means we can make a web page where other people can try out your model! View it here: https://replicate.ai/huage001/adaattn

Do claim your page here so you can own the page, customize the Example gallery as you like, and we'll feature it on our website and tweet about it too.

In case you're wondering who I am, I'm from Replicate, where we're trying to make machine learning reproducible. We got frustrated that we couldn't run all the really interesting ML work being done. So, we're going round implementing models we like. 😊

Hi, thanks for the repo and for the video test code posted in the issues! It would be extremely helpful for the video training script to be posted as well, so we can experiment training on specific datasets.

Thanks anyways

Hello, I have tried using the pre-training model you provided, and the effect is very good. Is the pre-training model you provided trained on Wikiart? If I want to train in more styles. Can you provide more details on how to train video style transfer? Just like readme said, is it OK to train directly on the coco dataset?