This paper received uniformly very positive reviews. All reviewers agreed that developing 3D vision systems that do not rely on 3D training data is an extremely important problem, that implicit surfaces are a promising representation, and that this paper proposes a useful technique that could inspire future work.