Critical infrastructure inspections are often carried out manually, which makes them costly, time-consuming, and often dangerous. A single event that shuts the facility down can cost a large operator $20–50 million USD in either loss production or deferred production. It can also have detrimental consequences for the environment. Abyss Solutions is on a mission to develop unique engineering solutions to replace manual inspections with automated ones.
Abyss Solutions uses V7 to label data and build AI-powered remote systems deployed to inspect large-scale infrastructure. The data collected is then processed to provide end-customers with insights that let them better manage and maintain their facilities. Abyss’ team is committed to lowering costs, mitigating risks, and ensuring a faster turnaround from inspection to action based on AI-powered insights.
Annotation systems are playing catch up trying to project where research and cutting-edge development would be and how to make it configurable. Luckily, solutions like V7 came up, which are maturing in the industry, helping companies like ours to scale up and stay ahead with our R&D
Abyss Solutions is an autonomous systems company delivering data-driven engineering assessments for critical infrastructure.
Founded in 2014 by four scientists and engineers from The University of Sydney, Abyss Solutions combines machine learning, computer vision, and robotics to develop AI-powered solutions for remote infrastructure inspections. Their work spans several industries—from the energy, water, and renewables sectors to agri-tech and more.
The company’s vision is to “build the intelligence that will drive all unmanned robots of the future” and thus replace current manual modes of infrastructure inspection—which are expensive and dangerous—with unmanned inspection systems.
Abyss Solutions is working with 2D/3D imagery, laser scans, and thermal data. The company has a dedicated robotics team whose task is to facilitate data capturing that later goes through the analytics pipeline and is delivered to end customers in a usable format.
Abyss inspects assets such as oil fields, inland refineries, and power plants, which are subject to adverse environmental conditions and are deteriorating over time.
The team labels captured data (e.g., corrosion, cracks, deformations) to train ML models that detect and segment objects from all viewpoints. Next, they merge all the insights and provide a consolidated view to the end customer—informing them about the health rating of a given asset compared to other objects.
Like many companies in the AI space, Abyss Solutions had to decide between building vs. buying their annotation software.
Prior to using V7, the team tested several labeling tools but quickly learned that the rapidly growing needs for more advanced functionalities could not be easily met. Abyss Solutions eventually settled for V7 as the fast feature requests turnaround helped them capitalize on new features and stay ahead with their R&D.
In addition, V7 enabled Abyss Solutions to collaborate with big teams—with up to 100 people on a single dataset—helping them complete their projects faster. The team could process tens of thousands of images in a month, update the model, and run quality control audits for their systems much more efficiently.
I like the auto-segmentation feature. To me, that’s a nice AI feature that V7 took beyond the gimmick feature - it’s mature enough to be useful.
Abyss Solutions aims to become a world-class player at inspection robotics by building computer vision-powered unmanned inspection systems. V7 was one of the factors that facilitated the success of Abyss’ data engineering team—helping it grow from 2 people to 100.
V7 helped Abyss scale their annotation process, manage terabytes of data more easily, and deploy their ML models faster. Abyss also started working on new projects—bringing more than five new types of data and interactivity with the data within V7 over the last 12 months.
Discover how other AI-first companies solved knowledge tasks at scale with V7