I know some techniques for augmenting data when images are used, but I don't know if there are any such techniques specifically catered to videos. Since videos have an additional temporal dimension compared to images I was wondering if there were any techniques that augments the video data using the view points and the motion information etc
You can augment videos in the temporal dimension through clipping, or taking random sequences of consecutive frames. You can also augment in the spatial dimension by cropping frames randomly to simulate camera jitter. Unfortunately, augmenting video data by extrapolating motion information or by producing a video taken from a different viewpoint seems extremely computationally intensive.