Leobit
AI-Powered Construction Progress Tracking App
Full-cycle software development allowing early detection of potential issues through AI analysis
ABOUT
the project
Client:
Location:
Norway
Company Size:
Industry:
Construction
Solution:
Services:
An AI-powered mobile application streamlines construction site management by providing real-time progress tracking, photo documentation, and predictive issue detection. This solution allows construction teams to proactively manage projects, detect issues, and minimize expenses on manual oversight.
We’re happy that the client trusted our advice and expertise and agreed to host the AI models on the backend. This decision allowed us to develop more robust models and significantly speed up image analysis.
Customer
Our customer is one of the leading construction companies in the Nordic region. They are involved in constructing residential, commercial, and public buildings, as well as infrastructure projects like roads, bridges, and tunnels. The company also invests in research and development to drive innovation in construction processes.
Business Challenge
The client needed a way to automate progress tracking and issue detection across multiple construction sites. Traditionally, foremen had to visit each site to evaluate progress. This manual process often led to delays, resource shortages, and overlooked quality issues.
Working in the construction industry for almost 20 years, the customer wanted to revolutionize this process by developing a solution that would allow workers to independently report on their progress and reduce the reliance on on-site visits by foremen. Additionally, they wanted to use AI to detect potential risks or structural defects risks.
Why Leobit
The customer discovered Leobit on Clutch and was impressed by the positive reviews and our expertise in AI integration and mobile development. They appreciated our attention to detail and the thoughtful suggestions we provided during the discovery phase, so they chose Leobit as their software development partner.
Project
in detail
Leobit developed a cross-platform mobile application using Flutter, which empowers construction workers to report on the progress of their projects directly from the field. The app provided an intuitive interface for workers to upload updates, add photo proofs of completed tasks, and request materials needed for upcoming construction stages. This removed the bottleneck of waiting for foremen to physically inspect every site, allowing for a more decentralized and efficient workflow.
Flow detection AI model development
During development, we created several AI models to identify issues such as wall cracks, misaligned corners, wall curvature, and other structural defects. These models were trained on large datasets that our client had provided. We used TensorFlow and PyTorch for model development and training. After extensive testing on a separate test dataset, the models achieved a 92.7% success rate.
Once the models were validated, we deployed them to Microsoft Azure’s GPU servers to take advantage of the faster performance of GPUs compared to CPUs. This allowed us to improve processing speed and optimize cloud costs. Deploying the models on the server also allowed seamless updates since users didn’t need to update the app whenever the AI models were improved or replaced.
Given that no AI model is perfect and might produce false positives or negatives, we built a feature into the app that allows foremen to manually review and flag issues. They can either confirm the AI’s detection or dismiss it, and these inputs help improve future model performance.
AR-assisted photo guidance
To improve the quality of photos taken by workers, we implemented AR-assisted photo guidance using the AR Flutter Plugin, which uses ARCore for Android and ARKit for iOS. This feature provides visual cues and hints to guide workers in capturing photos at the correct angle and position.
Thanks to this, the photos workers take, meet the necessary standards for AI analysis. Without this assistance, photos taken at incorrect angles could be flagged by the AI as unusable or misinterpreted, which could lead to inaccurate results. The AR ensures that workers capture accurate and clear images, making the AI analysis more reliable and reducing the risk of false results.
Caching and background services
Since construction sites often have poor internet connectivity, we utilized the Drift library (SQLite) for local data caching. This allowed us to temporarily store project photos on the device until a stable connection was available. By caching photos locally, workers could continue documenting progress without delays, even in low-coverage areas.
Once the worker took the photos, they were archived and automatically uploaded to the server when connectivity improved. To support this, we implemented background services that trigger the upload process only when a stable internet connection is detected.
Once uploaded and processed, the construction company receives a push notification with a detailed report or an alert if any issues are detected.
Real-time progress updates
The app allows workers to log task completion and submit status reports in real-time, providing a live feed of progress directly from the construction site. This decentralized reporting system allows foremen and project managers to monitor multiple sites remotely. To enhance this functionality, we implemented real-time photo uploads using Flutter for the front end and Azure Blob Storage for secure cloud storage and instant accessibility. We also used Microsoft Azure for cloud-based data synchronization to ensure seamless updates across devices.
Technology Solutions
- AI model development using TensorFlow and PyTorch, which achieved a 92.7% accuracy rate in detecting wall cracks, misaligned corners, wall curvature, and other structural defects
- AR Flutter Plugin to provide visual guidance to workers in capturing photos at the correct angle and position
- BLoC architecture implementation, which made the app more maintainable and responsive to real-time updates
Value Delivered
- AI-powered defect analysis significantly reduced the need for frequent on-site visits by foremen
- Cutting company’s operational costs while maintaining high-quality oversight
- Intuitive UI/UX design and user flow