Semantic Structure from Motion for Railroad Bridges Using Deep Learning
Round 1
Reviewer 1 Report
The authors propose through this work an automated method to recognize the structural components of railroad bridges and then construct semantic 3D models using photographic data collected by an unmanned aerial vehicle. A case study was conducted, and precision and accuracy of the method were measured.
The paper is well presented and structured but in my opinion, has two major issues:
+ State of the art should be improved especially on image classification/segmentation and form recognition using deep learning. Sentences like the ones below should be referenced and linked to the existing research works:
Lines 83 & 84: "Computer vision, a field that enables computers to recognize and analyze visual information, has been continuously developed [REFS??]. Deep learning techniques have recently introduced numerous applications in computer vision [REFS??]".
+ The proposed method and results should be discussed/ compared according to relevant related works on image processing and automatic 3d model construction. A discussion section would be a real + to this work. For example, what difference/similarity/complementarity in terms of process, precision, accuracy... between this work and the one introduced in [4]?
Author Response
We appreciate the reviewer’s valuable comments for improving the quality of the manuscript. We have carefully addressed each comment and revised the manuscript accordingly. Once again, thank you for the valuable comments of the reviewers.
Author Response File: 
Author Response.docx
Reviewer 2 Report
- page 3 Figure 1: structure of the CNN is really very concise; it doesn't hint at anything;
- rows 181-182 <<… while preserving the information on bridge components.>>: can you give, here in advance, some examples of preserved informations?
- page 10 Eq. 4, 5, 6, 7 and 8: find a way to better format the punctuation marks at the end of the equations, they look like quotes of the terms in the denominator;
- page 10 row 184 <<BF-score>>: define this acronym as it is mentioned for the first time;
- other references about infrastructural BIM: https://doi.org/10.3846/enviro.2020.683; BIM parametric modelling of a railway underpass, Ingegneria Ferroviaria, 2020, 6: 443-459; https://doi.org/10.3390/infrastructures5050041;
- congratulations, very well done job.
Author Response
We appreciate the reviewer’s valuable comments for improving the quality of the manuscript. We have carefully addressed each comment and revised the manuscript accordingly. Once again, thank you for the valuable comments of the reviewers.
Author Response File: Author Response.docx
