vipermu/ganimation

Pytorch implementation of GANimation with pretrained weights.

ganimation's Issues

in the wild images examples

can you show some examples of tests on images outside the dataset?

How to run if i only have the attribute_txt but not the attribute images?.

Essentially only the AU's are required according to the paper. However if i empty the attribute_images folder, the code wont run. Can you clarify if the attribute_images are made available only for Output comparison or are used in the code too?

Why are only AU[2]-AU[19] considered out of the 46 AUs?

Cannot open the links

Hi, thank you for your contribution to ganimation project. However, I could not open the links of your pretrained models. Could you please check the links again?

Is there any extra necessity I should convert these images to GIF?

Sorry, repeat submit, please delete this issue, thanks.

Deal with multiple faces

Great Work! Your work is of great help for me.
But I have a question now, can your model handle multiple faces ? If I have a picture that contains multiple faces, how can I create a face animation?
Waiting for your help,thanks

RuntimeError: CUDA error: unknown error

Hello! Thanks for your work.
But I have a problem while using "python main.py --mode animation"
I test with cuda9.0 cuda10.0 and cuda10.1 but result in the same error, how can I solve this problem?

Training images: 156295
Testing images: 100
Model running on cuda:0
Traceback (most recent call last):
File "main.py", line 60, in
main(config)
File "main.py", line 25, in main
solver = Solver(data_loader, config_dict)
File "C:\Users\15974\Desktop\ganimation-master\solver.py", line 35, in init
self.build_model()
File "C:\Users\15974\Desktop\ganimation-master\utils.py", line 17, in build_model
self.g_repeat_num).to(self.device)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\nn\modules\module.py", line 386, in to
return self._apply(convert)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\nn\modules\module.py", line 193, in _apply
module._apply(fn)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\nn\modules\module.py", line 193, in _apply
module._apply(fn)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\nn\modules\module.py", line 199, in apply
param.data = fn(param.data)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\nn\modules\module.py", line 384, in convert
return t.to(device, dtype if t.is_floating_point() else None, non_blocking)
File "D:\tools\anaconda3\envs\gan\lib\site-packages\torch\cuda_init.py", line 163, in _lazy_init
torch._C._cuda_init()
RuntimeError: CUDA error: unknown error

Generate attribute_image

Great Work! Your work is of great help to me. But I have a problem.
How can I generate attributes.txt in folder eric_andree?

Please help me
Thanks！

Error on using pretrained models on CPU only machine

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location='cpu' to map your storages to the CPU.
Even after changing the map_location to 'cpu' the problem still persists and ask for a GPU.
Any help is appreciated.

Could you tell me how many images were used as the data set?

I'm going to train a new model on CelebA-HQ with a pixel size of 384*384px to test the effect of high quality faces, have you tried it? Too large data set will take too long and it is more difficult to train with such large pixels, could you tell me how many images were used as the data set?

so hard to train

Hi, I use the preprocessed CelebA dataset that you provided to train, but the attention masks become 1 after several iters. I didn't change the parameters in the code.
lambda_cls=160, lambda_rec=10, lambda_gp=10, lambda_sat=0.1, lambda_smooth=1e-4.
When I change lambda_sat to 1, the attention masks also saturate to 1.
But when I change lambda_sat to 1.5, the attention masks become 0.
Your pretrained model works well，did you use the same parameters？
Thanks！

Missing action unit

Hello @vipermu, thanks for your great work.

I have a question regarding action units on attributes.txt.
OpenFace is able to recognise 18 AUs, but attribtues.txt seems to only have 17 AUs. Could you please inform me which AU is missing?

Is there any extra procedure I should convert these images to GIF?

Hi, your project is brillant and I have recurrent it on my machine.

I have a question that how could I got the result like GIF in your readme? Because the results in local result folder are image, is there any extra procedure I should convert these images to GIF?

Thanks!