Long Short View Feature Decomposition via Contrastive Video Representation
                 Learning
       Nadine Behrmann1    Mohsen Fayyaz2,     Juergen Gall2    Mehdi Noroozi1
           1                       2
            Bosch Center for Artificial Intelligence    University of Bonn
      {nadine.behrmann, mehdi.noroozi}@de.bosch.com, {fayyaz, gall}@iai.uni-bonn.de


               Abstract
  Self-supervised video representation methods typically
focus on the representation of temporal  ...