Fetching the paper…

DistInit: Learning Video Representations Without a Single Labeled Video · Around