Training data about ganimation HOT 17 CLOSED

Sorry, I am not the dataset owner. I only have a license to use it, but not to share it. I'm sure they will answer you in a few days.

from ganimation.

Hi, thank you very much for sharing your training and testing codes.
Could you please also share the training data?
I have applied for the training data on the website but still get no reply.
Thank you very much.

Hello, I am also waiting for the reply for months. I am so sad. Do you have the dataset now? Could you share with me? Thanks!
My email: [email protected]

from ganimation.

@HuayueZhang Hi, have you got the dataset now?

from ganimation.

@staceycy Could you share the training dataset with me? Thanks!
My email: [email protected]

from ganimation.

@IvyGongoogle Sorry, I didn't get the data. :(

from ganimation.

it is a sad thing

from ganimation.

You can find preprocessed data and pretrained models here: https://github.com/vipermu/ganimation

from ganimation.

You can find preprocessed data and pretrained models here: https://github.com/vipermu/ganimation

it is only the dataset celeba, but can not find preprocessed data about EmotionNet

from ganimation.

You can find preprocessed data and pretrained models here: https://github.com/vipermu/ganimation

it is only the dataset celeba, but can not find preprocessed data about EmotionNet

It does the job. EmotionNet is not a public dataset.

from ganimation.

I asked for dataset but I have not received it, it seems very hard to obtain it

from ganimation.

I asked for dataset but I have not received it, it seems very hard to obtain it

You can get pretty good results with celeba. I think that the only cool thing about emotionNet is the number of samples and high variability that it presents, but you need to crop the face region and re-compute the AUs anyway, so any dataset with faces should work. You can just find other datasets or create your own one and do the required preprocessing. I uploaded the preprocessed celeba to avoid all that stuff though.

from ganimation.

yeah probably you're right, but Emotionnet has a huge number of manually annotated images!

from ganimation.

@vipermu training without efficient epochs on emotionNet seems not to have a good performance. Like I get the problem. So emotionNet dataset is necessary?

from ganimation.

@vipermu training without efficient epochs on emotionNet seems not to have a good performance. Like I get the problem. So emotionNet dataset is necessary?

What do you mean by efficient epochs? I think that if you get enough data with enough variance on it the performance of the model will be the same.

from ganimation.

@vipermu Thanks for your reply. So celeba is enough data with enough variance?

from ganimation.

@vipermu Thanks for your reply. So celeba is enough data with enough variance?

Unfortunately, the variance in the images is really low but I think that there are enough samples to have something working. The performance won't be the same as the results on the paper, but you can get some decent results.

from ganimation.

@vipermu Thanks for your reply. So celeba is enough data with enough variance?

These results were generated with celeba: https://github.com/vipermu/ganimation/blob/master/video_results/standard_celeba.avi

from ganimation.