Dataset of 10000 real videos?

[jumpycat]

hi, thanks for the nice work, how can I get the 10000 real videos set(the selected and processed frames in VoxCeleb1)?
And also, is there an official Train/Val/Test split protocol?

[huangjiadidi]

I have update the dataset which include the real data.

For fake data, we select the first 750 videos in each class as the train data. The next 150 videos as val and the last 150 as test.

Thank you

[jumpycat]

I have update the dataset which include the real data.

For fake data, we select the first 750 videos in each class as the train data. The next 150 videos as val and the last 150 as test.

Thank you

Thank you for your work.