Deep Learning-Based Traffic Characterization for Smart Cities (Phase 1)

This project is situated within the context of Smart Cities, aiming to enhance urban quality of life through the application of advanced technologies. The problem addressed is the need for precise and automated traffic characterization. The proposed solution focuses on utilizing Deep Learning (DL) for vehicle identification within aerial images captured by drones. The methodology involves studying the state-of-the-art, acquiring imagery via drone photogrammetry, manually labeling vehicles in the generated stitched maps, and subsequently training and validating a DL model for object recognition. Once vehicles are identified, key traffic characterization metrics can be derived, such as traffic counts, speed measurements, congestion identification, and pedestrian/bicycle counts. This approach transforms visual data into actionable information for intelligent mobility management systems.<br/><br/><b>Goal</b>: <br/>The main objective of this project is to characterize traffic conditions around the UPS South Campus by applying Deep Learning techniques for the identification and analysis of vehicles in aerial imagery.<br/><br/><b>Research lines</b>: <br/>Geospatial sciences