Challenge:
The client is a leading global engineering firm specializing in the assessment and rehabilitation of underground sewerage systems. With a vision to revolutionize the way inspections are conducted, they sought to develop a cutting-edge Computer Vision-based program that could automatically identify defects within sewer pipes. The program would allow the user to easily go through the video frames and identify defects, and their clock positions, and allows users to see and interact with the web interface.
In the past, the firm relied on PACP-certified experts manually marking defects occurrence for evaluating the condition of gravity sewers, a process that was not only labor-intensive but also highly susceptible to human error. This approach relied on the interpretation of individual inspectors and their ability to identify defects within the pipes, leading to inconsistencies in assessment quality. By seeking an innovative solution to automate the inspection and quality review process, the client aimed to raise the bar and set new standards in the field.
Solution:
To cater to the client’s requirements, our team devised a comprehensive plan to foster a data-centric approach without relying on any field tools. The plan involved delivering the following:
- A web interface to extract and process a list of all the completed and ongoing projects of the contractor in the city. The system will support additional cities in the future.
- An interface for uploading 100+ videos to the server collected by the contractor.
- Implementation of the Computer vision-based Defect Classification algorithm to analyze every uploaded video for defects, and subsequently upload the identified issues.
iLink helped create an AI video processing algorithm for the detection and classification of sewer defects using Deep Learning and Computer Vision. This innovative product was recognized with the Digital Products Award at the Innovation Spotlight Awards 2022.
Process:
The project was executed in three phases:
- Phase 1: In the initial phase, iLink focused on training the AI engine for Sewer Pipe Defects Detection. The team used gigabytes of images to train the engine, ensuring that it could identify defects accurately. To facilitate easy integration, the Web App and AI engine were synced up as Microservices. Once the engine was ready, it was tested and found to allow contractors to identify defects more quickly and efficiently.
- Phase 2: The second phase of the project focused on improving the accuracy of the AI engine and scaling the Web App. To achieve this, the iLink team carried out intensive Data QA| QC. They also implemented state-of-the-art models to ensure that the engine could identify defects with even greater accuracy. To facilitate scaling, the team integrated Azure ML and ACI, allowing the system to handle increased load and perform at a higher capacity.
- Phase 3: In the final phase of the project, the Web App was scaled up, and the AI engine was further improved. The iLink team implemented a new feature known as Clock Position and Distance Detection of Defects, which enabled contractors to identify the precise location and extent of a defect. This resulted in a significant productivity improvement, with contractors able to identify defects four times faster than before. The new features also resulted in increased productivity (4x time faster), allowing contractors to complete projects more quickly and efficiently.
Results:
By processing 10,000 minutes of video review, the platform enabled a four-fold increase in productivity. This allowed contractors to identify defects in the sewer inspection media more quickly and efficiently, saving valuable time and resources.
In the initial testing phase, the platform demonstrated an accuracy rate of 60%, resulting in a 1.5 times increase in productivity. However, through continued development, the AI engine was further improved, achieving an impressive accuracy rate of 85% in the final phase of the project.
Overall, the platform had a transformative effect on the client’s organization, becoming a foundation for other entities to build upon. With the ability to differentiate between various types of defects accurately, the platform provided enhanced accuracy and productivity, improving the efficiency of the entire sewer inspection process.
