HKU transforms urban safety with 'eCheckGo' AI system for rapid building inspections
Assessing and undertaking building maintenance in densely populated cities like Hong Kong faces growing challenges as ageing buildings become more common. Manual inspections are often time-consuming and costly, leading to delays in identifying structural defects that can compromise public safety. To address these issues, a team in iLab, The University of Hong Kong (HKU), led by Professor Junjie Chen and Professor Wilson Lu from the Department of Real Estate and Construction, Faculty of Architecture, has developed eCheckGo, an innovative artificial intelligence (AI) system powered by a proprietary Large Defect Model (LdM) combined with traditional AI algorithms.
This technology recently received the Gold Medal with Congratulations of the Jury at the 51st International Exhibition of Inventions in Geneva, recognising its scientific excellence and significant potential for real-world application.
Breaking Barriers: Speed, Simplicity and Scalability
The risks associated with building defects are only set to increase as Hong Kong's building stock continues to age. At the end of 2020, the government estimated that around 8,700 privately owned buildings were aged 50 or above; by 2030, this number is expected to rise to nearly 14,000. Staying on top of maintenance and repair needs has therefore become increasingly problematic, particularly as a single building inspection can take several days to complete using conventional methods.
The eCheckGo system redefines traditional building inspections by delivering rapid, scalable, and adaptable assessments. In contrast to labour-intensive traditional methods, eCheckGo can analyse dozens of images within seconds, performing at least 100 times faster and eight times more cost-effectively than other automated technological solutions. This significant efficiency gain enables more frequent, proactive monitoring of building conditions at scale.
These capabilities are powered by eCheckGo's proprietary Large Defect Model (LdM), a large multi-modal model that has been trained on internet-scale datasets for building inspection. The model has been fed domain-specific inspection images with textual prompts, enabling consistent, reliable defect identification across diverse building types.
Designed for ease of use, eCheckGo allows users to capture images inside and outside buildings using a mobile app, or to leverage readily available Google Street View images. The AI system automatically detects defects such as cracks and spalling where concrete breaks away, and integrates these findings into generated 3D datapoint clouds. This interactive 3D model allows users to zoom in and out to view the exact geometry and dimensions of the issues.
Proven Performance and Future Impact
To further assess the system's large-scale capabilities, the team leveraged existing secondary data to conduct a city‑wide test using Google Street View images of 9,172 buildings in Kowloon. In just four hours, eCheckGo generated a colour-coded map grading building defects on a scale from 0 (healthy) to 10 (dangerous). These results were then verified against on-site inspections by professional building surveyors.
Highlighting the importance of early detection, Professor Chen said, "Building inspection is crucial for identifying defects and taking remedial action as quickly as possible, but in reality, it is often not done sufficiently because it is labour-intensive and time-consuming." He added that the strength of eCheckGo lies in its ability to present a comprehensive picture of building conditions within an intuitive 3D environment, saying, "You can easily understand the overall condition of a building, pinpoint where defects are located, and assess their scale, geometry, and dimensions. Having all this information consolidated in one place is extremely helpful for making informed and timely decisions about maintenance or renovation."
From Research to Community Impact
eCheckGo has attracted interest from government bodies and industry groups, with discussions underway for community adoption. The research team is currently working to expand the system's capabilities to detect water leakage and dampness, as well as generate automated text reports that align with professional formats. This initiative exemplifies HKU's commitment to translating cutting-edge research into practical solutions that address urban decay in high-density cities like Hong Kong, Tokyo, and Singapore.
