Before
Manual process to inspect sewer pipelines to log defects using a camera on a bogie. Very time consuming and limited coverage.
After
Computer vision based fault detection, automating the inspection process and increasing coverage by over 3 fold.
The How
- Understand the inspection process and camera data captured and how it is labelled
- Data Engineering phase to improve data sets
- Develop a transfer learning model for image classification
- Develop a Sewer fault detection tool to help label videos correctly and quickly
- Provide alerts on the exact locations of blockages, cracks and intrusions in the network
