Reviews: Combinatorial Energy Learning for Image Segmentation

The paper is well-written, with lots of details provided. The studied problem is definitely of great impact and would significantly benefit the community of neuroscience. The reviewer noticed the large size of the problem and appreciates the considerable amount of works achieved, as described by the paper. On the other hand, I also want to point out some associated weakness: 1. To my understanding, two aspects which are the keys to the segmentation performance are: (1) The local DNN evaluation of shape descriptors in terms of energy, and (2) The back-end guidance of (super)voxel agglomeration.