Open AI system forecasts construction delays using on-site cameras and data

A new open reference architecture proposes using AI to forecast schedule slips at construction sites. The system ingests data from multiple sources like crane logs and GPS into a central Kafka backbone, creating a unified object model. It employs edge devices with computer vision for real-time detection and a large language model for forecasting and planning. The architecture is designed for a large project in Saudi Arabia, addressing challenges like high temperatures and multilingual workforces. All components, including the model weights and object definitions, are released under permissive open-source licenses.