Trained on Massive Amounts of Data
At the heart of our technology, deep neural networks are responsible for reliably detecting, localizing, and quantifying all defects in your sewer pipes. Our models are capable of distinguishing more than 300 main- and subclasses of defects, which enables the reporting of all damages required by the European, British, and US coding standards. Our neural networks are trained on massive amounts of data: millions of images gathered in hundreds of different cities around the world.
Detailed 3D Reconstructions
To accurately locate each detected defect in the real-world, we build a detailed 3D reconstruction of each pipe. Our state-of-the-art computer vision pipeline tracks hundreds of thousands of keypoints in pipe images to recover the position of the robot and the shape and structure of the pipe with millimeter precision.
To get the best possible view of the pipe, we have developed algorithms to automatically select the regions of the image with the best exposure and sharpness and crop out water drops on the lens.
High Availability & Reliability
To guarantee 24/7 availability and maximum reliability, our software has to pass thousands of unit, integration, and end-to-end tests in a simulated environment before being released to our customers. Modern engineering practices allow us to develop new features quickly and deploy new releases within hours.