Master of Science in Computer Science
First Committee Member
Second Committee Member
Third Committee Member
Fourth Committee Member
Venkata Prashant Modekurthy
Fifth Committee Member
In the past few years, computer vision has made huge jumps due to deep learning which leverages increased computational power and access to data. The computer vision community has also embraced transparency to accelerate research progress by sharing open datasets and open source code. Access to large scale datasets and benchmark challenges propelled and opened the field. The autonomous vehicle community is a prime example. While there has been significant growth in the automotive vision community, not much has been done in the rail domain. Traditional rail inspection methods require special trains that are run during down time, have sensitive sensing/imaging analysis equipment with high costs, or may require low speeds for analysis. In addition, the lack of available labeled datasets for the rail domain has limited progress in the field. In this thesis, we explored and evaluated machine learning algorithms for real-time railroad inspection systems from the ego-perspective of the locomotive. This was accomplished through a study on state-of-the-art semantic segmentation models on popular automotive datasets with semantic segmentation annotations. Second, transfer learning was performed on the models with a public rail dataset. Third, benchmarking was done on the newly trained rail models on an embedded system and PC. Finally, a custom dataset was created to highlight anomalies on rails (e.g., mud pumping and vegetation).
Anomaly Detection; Computer Vision; Machine Learning; Railway; Real-Time Semantic Segmentation; Semantic Segmentation
University of Nevada, Las Vegas
Stanik Iii, Paul, "Real–Time Semantic Segmentation for Railway Anomalies Analysis" (2022). UNLV Theses, Dissertations, Professional Papers, and Capstones. 4623.
IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/