Periodic Reporting for period 1 - Twinspect (First fast and automated (<5 min) AI-based inspection to increase operational safety of large critical infrastructure, e.g., bridges, dams, and oil & gas platforms and refineries)
Okres sprawozdawczy: 2023-12-01 do 2024-11-30
Critical infrastructure, like the 1,234 km of <100m long road bridges in the Trans-European Transport Network, must be inspected every 3 years. But inspections are expensive, ~€4.9k/bridge: amortised €550k CAPEX, and €37k OPEX/vehicle, plus salaries for inspection engineers, and €1,400 costs for closing the infrastructure during inspection. Drones reduce the total inspections 77% on average to €1.1k/ bridge: 98% reduction in CAPEX and 88% in OPEX, as well as 100% savings from inspecting the infrastructure without disrupting its normal operation. However, drone inspections require a large amount (<2,000) of high-resolution (<20MP) photos to be taken from different orientations, resulting in large project size (<30 GB of data). Therefore, data storage and analysis costs avg. €912/bridge are the key bottleneck. Automated AI-based systems, like Twinspect, can further reduce 65% of total inspection costs (92% cost savings compared to on-site direct inspection by engineers).
Twinspect is the first AI-based photogrammetry solution that enables engineers to navigate and inspect highly fluid (>60 fps) and photorealistic (<50 GPX texture level detail and <200M polygons mesh density) 3D models of the infrastructure directly. It is the first and only solution that can be used for large infrastructure (>100k photos/project), e.g. bridges, dams, oil & gas extraction platforms and refineries. Our minimum viable product (MVP) consists of client-based photorealistic fluid 3D model digital twin technology, with automated AI-based defect clustering, detection, and annotation, and is already available (100+ paid pilot projects) as a SaaS to infrastructure owners directly, or via inspection services providers, e.g. engineering or drone companies. During the project we aim to enable, for the first time, the overlap of two 3D models from consecutive inspections for seamless visual inspection and develop AI tools to flag new or growing defects.
Twinspect enables more accurate (100% defect detection), consistent (collaborative annotation and ranking), and fully documented inspections (clustered photos and annotations). It also provides a complete record and full audit trail for the first time. This is a key advantage of first to market, that is to gather high-quality historical data, that we will leverage to develop deterioration models and predictive analytics tools to enable advanced asset management.
Twinspect is the first AI-based photogrammetry solution that enables engineers to navigate and inspect highly fluid (>60 fps) and photorealistic (<50 GPX texture level detail and <200M polygons mesh density) 3D models of the infrastructure directly. It is the first and only solution that can be used for large infrastructure (>100k photos/project), e.g. bridges, dams, oil & gas extraction platforms and refineries. Our minimum viable product (MVP) consists of client-based photorealistic fluid 3D model digital twin technology, with automated AI-based defect clustering, detection, and annotation, and is already available (100+ paid pilot projects) as a SaaS to infrastructure owners directly, or via inspection services providers, e.g. engineering or drone companies. During the project we aim to enable, for the first time, the overlap of two 3D models from consecutive inspections for seamless visual inspection and develop AI tools to flag new or growing defects.
Twinspect enables more accurate (100% defect detection), consistent (collaborative annotation and ranking), and fully documented inspections (clustered photos and annotations). It also provides a complete record and full audit trail for the first time. This is a key advantage of first to market, that is to gather high-quality historical data, that we will leverage to develop deterioration models and predictive analytics tools to enable advanced asset management.
During the first year of the project, we have performed the following activities related to the product development:
- We conducted high profile demonstrators of large critical infrastructure, collecting performance data in real operational conditions and end-user’s feedback.
- We developed and refining methods of collecting performance data and benchmarking against competitive solutions.
- We successfully achieved ISO 27001 certification in May 2024, ensuring our commitment to information security standards.
- We completed the development of algorithms to automate training data selection.
- We are implementing an active learning system for our AI damage detection which continuously incorporates new user feedback and additional data sources to enhance the AI's performance and accuracy.
- We have experimented with hyperparameters to enhance texture quality following the optimizations made for client-based rendering of the digital twin.
- We have completed the development of the automation of the 3D model and texture optimization.
- We have developed a quality test system with a computer vision AI engine to compare 3D model rendered images against the original high-resolution (drone) photos.
- We held major releases in 2024, each introducing key improved features to our technology to enhance user experience and functionality.
- We have signed resellers by now eleven resellers’ agreements already secured in different countries such as USA, Mexico and Japan among others.
- We conducted high profile demonstrators of large critical infrastructure, collecting performance data in real operational conditions and end-user’s feedback.
- We developed and refining methods of collecting performance data and benchmarking against competitive solutions.
- We successfully achieved ISO 27001 certification in May 2024, ensuring our commitment to information security standards.
- We completed the development of algorithms to automate training data selection.
- We are implementing an active learning system for our AI damage detection which continuously incorporates new user feedback and additional data sources to enhance the AI's performance and accuracy.
- We have experimented with hyperparameters to enhance texture quality following the optimizations made for client-based rendering of the digital twin.
- We have completed the development of the automation of the 3D model and texture optimization.
- We have developed a quality test system with a computer vision AI engine to compare 3D model rendered images against the original high-resolution (drone) photos.
- We held major releases in 2024, each introducing key improved features to our technology to enhance user experience and functionality.
- We have signed resellers by now eleven resellers’ agreements already secured in different countries such as USA, Mexico and Japan among others.
During the first year of the project, our solution has improved key features to enhance user experience and functionality such as georeferenced 3D models, customizable project settings, multi-language support, and an improved report overview. Our solution has the potential to reduce 65% of total inspection costs (92% cost savings compared to on-site direct inspection by engineers) offering speed as well as high accuracy. Timely and accuracy information from inspections is key for infrastructure owners and operators to service the damaged infrastructure, which plays a key role for a country’s economy and the integration of Europe.
Our solution has an important impact for the region since it contributes to strength Europe’s position in the digital world allowing it to be pioneer in the digital-inspection of infrastructure as we provide the first AI-based solution for the efficient and at the same time precise remote inspection of critical infrastructure, solving current pain points like high costs, long inspection processes, and a lack of specialists for infrastructure inspection. Our solution has the potential to go global; thus, strengthen the European role in this market. Parallelly, we contribute to the “European Green Deal” by promoting early detection and preventing infrastructure damages and that can result in harmful consequences for the natural environment, for example preventing of leaks at plants of oil & gas companies which can end in catastrophic events.
Our solution has an important impact for the region since it contributes to strength Europe’s position in the digital world allowing it to be pioneer in the digital-inspection of infrastructure as we provide the first AI-based solution for the efficient and at the same time precise remote inspection of critical infrastructure, solving current pain points like high costs, long inspection processes, and a lack of specialists for infrastructure inspection. Our solution has the potential to go global; thus, strengthen the European role in this market. Parallelly, we contribute to the “European Green Deal” by promoting early detection and preventing infrastructure damages and that can result in harmful consequences for the natural environment, for example preventing of leaks at plants of oil & gas companies which can end in catastrophic events.