1 dataset found

The dataset used in the paper is not explicitly described, but it is mentioned that the authors collect 3K videos (60-90 seconds long) from the internet to train their model.