We’re excited to share that our collaborative team from the Automotive Innovation Lab at STU has just published a comprehensive camera and telemetry dataset featuring nearly 33 hours of real-world driving across diverse Slovak roads in Scientific Data, a prestigious open-access journal from Nature Portfolio (nature.com). Known for its rigorous peer review and commitment to high-quality data descriptors, Scientific Data ensures that this dataset is robust, reproducible, and valuable for the wider research community.
A Rich, Real-World Resource for AI and V2X Research
Recorded with an automotive-grade FPD-Link camera at 30 fps, high-precision GNSS (with ~14 mm accuracy), and LTE/5G telemetry, the dataset spans city streets, highways, and rural roads under lighting and weather conditions ranging from bright daylight to dawn, rain, and fog (nature.com). Over 3.5 million synchronized data samples—including position, velocity, heading, and network connectivity—provide an excellent foundation for computer vision and connected vehicle research. Tasks like lane and road-sign detection, visibility analysis, and mobile network performance benchmarking are now better supported by this real-world resource.
Reliable environment perception lies at the heart of advanced driver-assistance systems and self-driving technologies. Our dataset offers continuous video streams, supporting temporal detection of traffic signs and lane features—ideal for training robust neural networks (nature.com). Meanwhile, the telemetry data enables studies on Vehicle‑to‑Everything (V2X) communications, aligning with 3GPP requirements and shedding light on EDGE cases like GNSS signal loss in tunnels or mobile network drops in rural areas. This dual-modality dataset offers a powerful tool to evaluate both visual and communication aspects of intelligent transportation.
Published under CC BY license on December 30, 2024, the AIL Albus dataset reflects our commitment to open science and transparency, especially since the work was publicly funded by the Slovak Ministry of Transport (nature.com). Accompanying pre-trained labels for lanes and road signs enable immediate use, and the entire dataset—1.7 TB across 96 drives—is easily downloadable from the AIL Lab portal. Moreover, anonymized license plates and faces respect privacy while empowering a wide range of non‑commercial research and student-driven projects.
Link to full paper: https://www.nature.com/articles/s41597-024-04276-y