Training data

[staceycy]

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.

[albertpumarola]

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.

[HuayueZhang]

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]

[imjiangjun]

@HuayueZhang Hi, have you got the dataset now?

[PapaMadeleine2022]

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

[staceycy]

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

[PapaMadeleine2022]

it is a sad thing

[vipermu]

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

[PapaMadeleine2022]

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

[vipermu]

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.

[bsmith0003]

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

[vipermu]

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.

[bsmith0003]

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

[PapaMadeleine2022]

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

[vipermu]

@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.

[PapaMadeleine2022]

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

[vipermu]

@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.

[vipermu]

@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