Questions for "DyStaB: Unsupervised Object Segmentation via Dynamic-Static Bootstrapping"
-----------------------------------------------------------------------------------------------

Please send your answers to: schroepp@cs.uni-freiburg.de by 13:30 on 31.01.2022.

Question 1. (~3 sentences) The presented DyStaB method involves a so-called dynamic model (phi), a conditional inpainting network (psi) and a static object model (chi). Describe the inputs and outputs of each model and the function of each model intuitively.
Question 2. (~2-3 sentences) The dynamic model is trained jointly with the conditional inpainting network psi. How does this joint training work and what is the idea behind it?
Question 3. (~2-3 sentences) The dynamic model and the static model are trained in a "synergistic loop" with a procedure called Dynamic-Static Bootstrapping. How does this training procedure work and what is the intuition behind it?