Scene Aligned Pooling for Complex Video Recognition

  Abstract