dex_imdb_wiki.caffemodel.pt

Question

dex_imdb_wiki.caffemodel.pt

XIAOKANGWANDE666 opened this issue 4 years ago · 13 comments

XIAOKANGWANDE666 commented 4 years ago

could you please provide the dex_imdb_wiki.caffemodel.pt?

XIAOKANGWANDE666 commented 4 years ago

thanks！

Answer 1 · 2020-09-18T05:52:16.000Z

Hi @XIAOKANGWANDE666

You can easily generate this model. The authors provide nice instructions (Credits to authors). It takes a few minutes.

But if you can't do that due to some reasons, download via this link. The file will be stored for 30 days.

Answer 2 · 2020-09-18T06:41:33.000Z

Thanks you very much!

Answer 3 · 2020-09-18T07:18:38.000Z

@XIAOKANGWANDE666, you are welcome!

If everything is fine, then you can close this issue to help authors.

Answer 4 · 2020-09-18T07:23:06.000Z

the dataset is so lange.Do you run the train code?I want to train at lower resolution picture,I don't know if it works?I will close it later.

Answer 5 · 2020-09-18T07:55:58.000Z

@XIAOKANGWANDE666

the dataset is so lange.

I am sorry, what do you mean by 'lange'?

Do you run the train code?

Yes, I ran it yesterday. I could train it, and see training results using tensorboard (saving to logs included in the code).

I want to train at lower resolution picture,I don't know if it works?

I used FFHQ @256x256 resolution for training, but I didn't fully train the network, just checked :) So, I am not sure whether it will perform well or worse. And you can see discussions related to this by checking other issue. By doing so, you can find ou that the authors trained the network for three different resolutions: 256x, 512x, 1024x.

Answer 6 · 2020-09-18T09:07:19.000Z

@XIAOKANGWANDE666, you are welcome! Good luck with your research

Answer 7 · 2020-09-18T12:17:34.000Z

excuse me？Do you train the code only with 70,000 1024x1024 pictures?I see that FFHQ has 70,000pictures of 1024x1024,but auther use styleGAN create more.The resdme only train 70,000?Coulad you please tell me how you do it?

Answer 8 · 2020-09-21T07:56:36.000Z

Hi, @XIAOKANGWANDE666

Sorry for the late reply.

I used the original FFHQ with 70K images along with labels provided by authors.

Yes, the authors used an augmented dataset, namely age-equalized, using StyleGAN (check Section 4.1 of the paper).
But unfortunately, I couldn't find any link to get this dataset. I even raised an issue here.

So, basically, you can run their code using the original set which I am doing these days.
But, I am not sure about performance by training using this set.

Answer 9 · 2020-09-21T08:04:13.000Z

how much time do you spend on training?

Answer 10 · 2021-03-19T14:38:01.000Z

Hi @XIAOKANGWANDE666

You can easily generate this model. The authors provide nice instructions (Credits to authors). It takes a few minutes.

But if you can't do that due to some reasons, download via this link. The file will be stored for 30 days.

Hi, could you please provide the .pt file for me as well? I had trouble when converting the caffe model

Answer 11 · 2023-09-04T03:47:49.000Z

could you please provide the dex_imdb_wiki.caffemodel.pt?

Answer 12 · 2023-09-13T02:28:30.000Z

你好

您可以轻松生成此模型。作者提供了很好的说明（作者的功劳）。这需要几分钟。

但是，如果由于某些原因无法执行此操作，请通过此链接下载。该文件将存储 30 天。

could you please provide the dex_imdb_wiki.caffemodel.pt?