Inconsistent amount in DFDC.

[wangty1]

Hi, thanks for the wonderful work.

As I was downloading and splitting the images according to the train/val/test partition, I happened to find that the DFDC images were much fewer than the number of items in the corresponding csv files. Is it common? Any comments on it?

By the way, there are left-over images for each of the four datasets after splitting train/val/test, is it common?

[Purdue-M2]

Hi,

We only used DFDC for testing. All items in test.csv are in the images.