I understand that there are 93k video clips with 13 frames at 2Hz.
But there is a way smaller amount of logs in the whole dataset. So my question is it possible to construct sequences that are longer than 13 frames for a specific camera channel. Or in other words, how can we get the longest possible continuous sequence for a specific camera channel.
As I understand now the keyframes with segmentation are not continuous in time but rather spread out. So it is not possible to get the segmentation for multiple keyframes in a row
Yes, I agree. Since the data was collected over several years, there may be some exceptions, but usually we try to space the annotated images out several seconds to get more diversity.