Use Case #1 · Lead: ALP.LAB · Université Gustave Eiffel

Enhanced logging & user-interaction monitoring for road infrastructure improvement.

Tackling urban road safety challenges through proactive monitoring of high-risk areas - fostering safer, more harmonious coexistence among all road users.

The Challenge

Urban roads weren't built for the mix of users they now serve.

Urban roads frequently experience high vehicle density, particularly at intersections where traffic converges, creating congestion hotspots. This buildup, combined with driver impatience, can often lead to risky behaviours and traffic violations - motorcycles and cars weaving through lanes or bypassing congestion in unsafe ways.

In densely populated areas, the lower vehicle speeds may foster a false sense of security. Many drivers and riders overlook fundamentals such as wearing helmets or fastening seatbelts, underestimating the risks. Urbanisation has also led to a rise in cyclists, e-bikers and pedestrians sharing roads designed primarily for motor vehicles.

The current infrastructure has struggled to adapt to the evolving mix of road users. Even a small lapse in attention can have severe consequences - particularly for vulnerable road users.

iDriving's response: proactive monitoring of high-risk areas and targeted interventions to enhance road safety, redefining urban road-safety standards for a better future.

Trial Summary · PUC 1.1 ✓ Completed Feb 2026

PUC 1.1 trials in Graz, Austria.

Between 16-20 February 2026, the consortium ran a five-day on-site trial in Graz - the first real-world validation of iDriving's enhanced logging and user-interaction monitoring stack. Testing kicked off at the Torus bus garage with physical setup and system integration, then progressed to operational scenarios with drones operated by Acceligence capturing traffic footage in tandem with the wider system.

5days
On-site trial duration (16-20 Feb 2026)
10classes
Detection categories: helmet, seatbelt, mobile, vehicle, plate, windshield + more
87% mAP
Average mean Average Precision of YOLOv11 models
11-12fps
Real-time inference on NVIDIA Jetson AGX Orin (25 FPS on GPU)
4,000images
Curated training dataset for the Graz scenario
UAV

Drone surveillance

UAVs operated by Acceligence captured traffic footage with on-board NVIDIA Jetson processing - object detection, incident detection and intelligence streaming back to the system.

CV

Real-time perception

YOLOv11 (Small and Medium) running on edge hardware with DeepSort and ByteTrack tracking - identifying helmet/seatbelt use, mobile distraction, vehicle classification and traffic violations.

OCR

Licence-plate recognition

An OCR pipeline integrated for licence-plate recognition under varied lighting - feeding the violation logging system in real time.

V2I

End-to-end integration

Confirmed data transmission, message handling and system integration across modules from CERTH, Université Gustave Eiffel, Tekniker and Netcompany-Intrasoft.

Expected impact

The numbers we're aiming for.

0%
Increase in helmet & seatbelt usage
0%
Reduction in minor & major collisions
0%
Decrease in peak-hour traffic congestion
0%
High-risk zones identified & rectified