About frame extraction from a dataset

[MacchiatoCN]

For the video extraction frames in the dataset, you don't seem to give information about the number of frames extracted for each video, what is your experimental setup in this regard?