Pipe Sleuth: AI-based pipeline assessment

Regular inspection of underground water and sewage pipelines helps prioritize periodic maintenance tasks that help avoid pipe leakage, breakage, and blockage, reducing property damage or safety hazards. National Association Of Sewer Service Companies’s (NASSCO’s) PACP is the North American standard for pipeline defect identification and assessment, providing standardization and consistency to the methods by which pipeline conditions are identified, evaluated and managed.

Videos of pipelines are generated by a camera-mounted rover maneuvered by an operator into the underground pipelines. The operator typically notes the observations and reports any problems discovered and their locations. The process of manually reviewing this data is time consuming and error prone. Wipro’s Pipe Sleuth is an approach to automate pipeline inspection from the recorded videos using techniques in computer vision and deep learning.

Sundar Varadarajan dives into Wipro’s Pipe Sleuth solution (in collaboration with its customer DC Water), which automates the process of identification of defects in videos, annotation of defects in the videos, scoring and grading of pipeline health, and generation of reports as per PACP standards. The deep learning and CV-based solution has been optimized for running on Intel platforms (CPUs) using Intel Xeon Scalable processors and Intel Distribution of OpenVINO toolkit for optimization of the solution for Intel CPUs. Sundar walks you through the Pipe Sleuth solution, which has been optimized to run on Intel platforms at acceptable levels of accuracy and performance.