**5. Conclusions**

To identify issues with the detection and visibility of pavement markings, relevant studies were reviewed. The automated condition analysis framework for pavement markings using machine learning technology was proposed. The framework has three modules: a data processing module, a pavement marking detection module, and a visibility analysis module. The framework was validated through a case study of pavement marking training data sets in the U.S. From the quantitative results in the experimental section, the precision of the pavement marking detection module was pretty high, which fully validates the e ffectiveness of the YOLOv3 framework. Meanwhile, observing the visual results, all the pavement markings are correctly detected with the rectangle boxes and classified with the attached text in the road-scene images. In addition, the visibility metric of pavement markings was defined and the visibility within the proposed framework was confirmed as an important factor of driver safety and maintenance. The computed visibility values were also attached besides the detected pavement markings in the images. If the proposed study is used properly, pavement markings can be detected accurately, and their visibility can be analyzed to quickly identify places with safety concerns.

From the distribution of the testing samples, it can be inferred that the proportions of the straight markings, the right straight markings, and the left straight markings could be very low. Enlarging and enriching the training dataset could be a goal for future research.

**Author Contributions:** Conceptualization, K.K., T.K. and J.K.; Data curation, K.K.; Formal analysis, K.K., D.C. and C.P.; Investigation, D.K. and J.K.; Methodology, K.K., D.K., T.K. and J.K.; Project administration, T.K.; Resources, T.K.; Software, K.K., D.C., C.P. and T.K.; Writing—review & editing, K.K. and T.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by a gran<sup>t</sup> (KICT 2020-0559) from the Remote Scan and Vision Platform Elementary Technology Development for Facility and Infrastructure Management funded by KICT (Korea Institute of Civil Engineering and Building Technology) and a gran<sup>t</sup> (20AUDP-B127891-04) from the Architecture and Urban Development Research Program funded by the Ministry of Land, Infrastructure, and Transport of the Korean government.

**Conflicts of Interest:** The authors declare no conflict of interest.
