The reviews were leaning positive, and after the rebuttal, two of the reviewers recommended to accept, while R3 remained at the "marginally below" score. I agree that most of the reviewers comments were addressed in the rebuttal, and concur with the majority vote here. The paper is an incremental, but worthy addition to the evolving semantic segmentation craft. I strongly encourage the authors to incorporate the responses in the rebuttal into the final version, and in particular, to make sure to include some qualitative results.