AI powered by Google’s VideoBERT can predict what will happen next in videos by learning visual-linguistic and visual representations from unlabeled videos.
Amazing self-supervised system that tackles various proxy tasks to learn temporal representations from unlabeled videos.
Read more at https://arxiv.org/pdf/1904.01766.pdf