5x5 Technologies - ML for digital twin assessment
Challenge 5x5 is a US-based startup focused on creating an innovative inspection and maintenance system for telecom masts. Their approach involves using a 3D digital twin of the masts to identify key characteristics, such as installed equipment and available space. As the company grows, scalability becomes crucial, particularly with plans to enter new markets. Our team joined the project to automate the augmentation of digital twin data and to provide a new system architecture for future development.
To significantly reduce costs associated with core business functions, we aimed to develop machine learning models that could automatically evaluate the 3D model of the mast, identifying and tagging its components, including specified objects, devices, and unused space. Our solution incorporates various techniques, ranging from deep learning for 3D object recognition to image analysis and 3D graph exploration of the digital twin. We plan to cut the time required for a complete analysis and annotation of the tower's digital twin by 60% with our implementations. Additionally, our engineers established a contemporary microservice-based architecture and a new data model to enhance the client's system development. With the introduction of more standardized coding practices, the system has become more reliable, stable, and manageable, as well as easier to expand.