AI Case Study
Skand reduces cost of asset inspection by 80% with machine learning
Skand, a Melbourne-based aerial asset inspection company, has leveraged Silverpond's technology to enhance inspection analysis. 3D visualisations, from aerial photographs, are constructed using to make detecting defects easier. Silverbrane, which is Silverpond's AI tool, is then trained using this aerial data to be able to recognise defects. The company has achieved an 80% reduction in asset inspection cost, 70% reduction in inspection time of the 6.2 million metres of roof, while 20,000 defects were detected.
Construction And Engineering
Silverpond and Skand worked together to develop an aerial asset intelligence product which uses a mix of human and machine intelligence to identify defects in buildings. 3D modelling software was used to construct 3D visualisations from aerial photographs.
Large buildings with uniform features can prove di cult for image detection in 2D photographs, so by creating
a 3D model of the building, defects can be detected much more easily.
The Silverpond team then used its AI tool, Silverbrane, to train the software to recognise defects by showing it hundreds of thousands of images. The more images that are used in the training phase, the better the likelihood that the software can recognise defects in a real-world application.
Skand inspectors were trained to identify errors and inconsistencies in the defects identified, creating an important feedback loop that improved the model with each image that was analysed."
"Traditionally, building inspections have been a labour intensive, costly and dangerous task, involving people accessing roofs and facades to identify defects. Drone technology has improved aspects of the process by enabling the ability to collect a large number of images of a building’s roof or facade safely.
However, there are issues with reliability, as an individual still needs to sift through all the images to identify and assess defects. And if the company has multiple sites and buildings, the results are more likely to be inconsistent because so many people are involved in assessing the imagery."
"In the 2017/18 nancial year alone, Skand have:
* Reduced the cost for the asset owner from $0.35/metre
* Decreased the inspection time by 70%
* Inspected 6.2 million metres of roof
* Removed the requirement for a rooftop inspection by a person for over 600 buildings
* Detected over 20,000 defects
* Increased the total coverage of a traditional inspection from 15% to 100% of the building’s surface area"
The company's AI tool, Silverbrane, has been leveraged "to train the software to recognise defects by showing it hundreds of thousands of images."