Video datasets and annotation at scale

Turn footage into labelled training data, including dense action annotation over time.