Monocular 3D Human Pose Estimation
---------------------------------------------------------------------------------------------------
Please send your answers to zimmermann@cs.uni-freiburg.de before 26.07.2017 10:00
---------------------------------------------------------------------------------------------------
1) In which scenarios does the fusion scheme help the 3D prediction according to the paper? (1 sentence)

2) Whats the advantage of the proposed skip connections over the regular way skip connections are used? Why do corrective skips decrease performance in some cases? (2-3 sentences)

3) How does taking an human centered image crop affect the 3D annotation and how does the paper deal with it? (3 sentences)