A

Most Videos are originally 1944x2592.To make the dataset collage as shown above, the original videos were downsampled and their speeds were matched. The underlying data has not been released yet! Please do not use the collages we preview as the final data to train ML models on.