Review Reports

Round 1

Reviewer 1 Report

• For weighted cross entropy, did you select a certain level or threshold for weights (wt).
• Need to be consistent in datasets representation. Fig 2 and fig6 and fig3. You do not have to display the entire training and test sample. You can just display samples if you decide to keep fig 3.
• A lot of figures to represent the results(16-27). This is quit distracted.
• Table(5) that you compared your  approach with reference[16] results are quit off. I you want to test your approach performance, it is a good to compared with more than one OR at least, the comparison should be in the same environment . 

Author Response

Please refer to attachment for the responce to your comments.

Author Response File: Author Response.docx

Reviewer 2 Report

This research developed a method to for building extraction using deep learning and in particular PointNet++.

The paper is fluid and well-structured: the first part of the paper starts with a brief introduction, followed by a review of state-of-the-art research on the topic. After that, the proposed method is deeply presented.  The workflow is well-defined and described and it presents promising results. However, the innovation and the potentialities of this research, in my opinion, are not well valorized and highlighted. My kind suggestion is to enhance the conclusions and to propose further applications also in other contexts.

Finally, I support the publication of this paper.

Author Response

Please refer to attachment for the responce to your comments.

Author Response File: Author Response.docx