construct the MD-FAS benchmark

[GSHZhang]

When I construct the FASMD dataset based on the list (SIW E train and OULU E train) you provided, I found that I got the E sub-dataset with 839(from siw)+1620(from oulu) videos, which is a lot more than the 1696 listed in the paper?

[CHELSEA234]

we have made changes on the benchmark in the paper, which contains governemnt data that is not allowed to share. We will update the arxiv.

[GSHZhang]

So, the benchmark provided here is the final version?

[CHELSEA234]

So, the benchmark provided here is the final version?
@GSHZhang
I am sorry for the last response. Yes, this is the final version.
In fact, many ids in three datasets are repeated. This is because I want to make each spoof attack balanced. As you can imagine, the balance between different spoof attack helps the converge.