CreativeBloke - Revolutionising remote 3D workflows with cloud workstation technology
Gaist
Learning to understand objects in images without supervision for robotics applications
PUBLISHED 22 MAY 2022
Gaist was founded in 2007, leveraging considerable knowledge in infrastructure surveying, to better understand the condition of highway infrastructure and assets and enable detailed lifecycle planning to be undertaken in a way that was not commonplace. Gaist has a reputation for developing highly intelligent systems that challenge legacy technology and methodologies, employing the next generation of AI within an award winning platform for organisations both in the UK and globally. These include local authorities, central government departments, software assets management companies, utilities suppliers, private infrastructure owners and highway maintenance contractors.
Project Background
The team at Gaist has always been passionate about utilising the power of data and machine learning technology to help it with its ambitious mission to build the world’s deepest, most sophisticated understanding of the roadscape environment. To that end, the company has its own AI Innovation Hub in North Yorkshire, where its own in-house computer scientists and research specialists work to deploy the latest deep learning research techniques to accelerate the development of tools, techniques and intellectual property.
Using its fleet of mapping vehicles, Gaist capture between 40-60 million images per month. Roof-mounted 360O 30 megapixel panoramic camera survey the environment every 2-5m, whilst carriageway 4K cameras capture road data every 1m.
Project Approach
In order to provide carriageway condition data of unprecedented accuracy and quality, Gaist requires powerful GPU-accelerated systems to train its models to recognise 35 damage types including potholes, cracks or uneven surfaces. It was decided GPU-accelerated workstations were needed to handle huge volumes of data in order to reduce bottlenecks in the analysis pipeline.
Following a testing period of various configurations, NVIDIA professional GPUs were chosen for their robust performance and high degree of support.
Project Results
Conducting image analysis using the GPU-accelerated systems proved key to reducing pipeline bottlenecks, ensuring Gaist could process data in a timely manner to keep pace with the influx of new images daily.
The Scan Partnership
Scan worked with Gaist to provide a variety of custom built GPU-accelerated systems designed by our in-house 3XS Systems division. These systems range from NVIDIA Data Science Workstations, powered by RTX A4000 & RTX A5000 GPUs for model development, and systems with Tesla T4 GPUs being used for inferencing. This GPU-accelerated hardware has helped form part of Gaist’s own AI Innovation Hub at its headquarters in North Yorkshire.
Project Wins
GPUs accelerated image identification and classification
Time and cost-savings generated due to rapid image recognition ability
Steve Birdsall
CEO, Gaist
Many companies collect data but few can turn that data into meaningful action. Our world-class team uses its deep understanding of our clients’ needs to deliver valuable insights that help streamline operations, reduce risk, and cut costs while increasing safety."
Speak to an expert
You’ve seen how Scan helped Gaist develop its road quality analysis solution. Contact our expert AI team to discuss your project requirements.
phone_iphone Phone: 01204 474210
mail Email: [email protected]