Object Detection at Construction Sites
Background
The construction industry is a vital part of our modern world, but it faces challenges due to various disruptions like weather, traffic congestion, breakdowns, and diseases. Even minor disruptions can cause costly delays, project cancellations, and increased emissions, affecting both the environment and project efficiency. Enter ECOLOGIC, short for Emission Control and Logistics Optimization for Green Infrastructure Construction, is on a mission to transform the Dutch construction logistics sector.
By harnessing the power of data-driven insights and advanced AI analysis techniques through a real-time Carbon Digital Twin. Imagine a virtual mirror that reflects the carbon footprint of construction operations, allowing us to anticipate and adapt to challenges even before they occur. The general approach consists of the "Triple A": Acquire (collect and analyze data), Anticipate (predict) and Adapt (adapt to situations).
You will have the opportunity to collaborate with some industry partners and real stakeholders, as this project is part of a larger sustainability-driven initiative ECOLOGIC (link).
Task
Develop a proof-of-concept AI-powered tool that will detect objects in video images captured on construction sites. This tool will enable to keep a watchful eye on the construction site, detecting various objects such as machinery and humans in a time-lapse series of photos. Deep learning approaches are expected to be used, including some manual annotation of training data in order to detect (a selection of) objects.
WORK
20% theory, 40% implementation (e.g., modelling and programming), 20% labelling/annotating, 20% writing
Contact:
Rob Bemthuis (r.h.bemthuis@utwente.nl)