Questions for "Hierarchical Long-term Video Prediction without Supervision"
---------------------------------------------------------------------------
Please send your answers to bechtolj@cs.uni-freiburg.de by 19.06.2019 10am


1) What is the advantage of a hierarchical method? How can the authors make use of it without having ground truth annotations? (~ 2-3 sentences)

2) The authors describe two alternative ways to train the networks. Name them and describe the difference. (~ 2-3 sentences)

3) What does the person detector evaluation show / demonstrate? (~ 2-3 sentences)