One Dataset. Every Weather. Better AI.
The Weather-nuScenes dataset, developed by researchers from the Information Processing and Telecommunications Center (IPTC) at Universidad Politécnica de Madrid (UPM) in collaboration with Technische Universität Berlin (TU Berlin), represents a significant contribution to the autonomous driving and trustworthy AI research communities. Created within the SMARTY Project, funded by the European Union’s Horizon Europe programme and the Chips Joint Undertaking (CJU), the dataset extends the widely adopted nuScenes benchmark by generating realistic adverse weather and nighttime conditions using state-of-the-art generative AI techniques.
Weather-nuScenes enriches existing driving scenes with synthetic rain, snow, and low-light environments while preserving the original scene geometry and semantic content. This enables researchers to evaluate and improve the robustness of computer vision and perception algorithms under challenging environmental conditions without the cost and complexity of collecting real-world data in hazardous scenarios.
By making the dataset openly available, the IPTC research team promotes reproducible research, accelerates the development of more resilient AI models, and provides an accessible benchmark for universities, research centers, and industry. The dataset is particularly valuable for training and validating perception systems based on cameras, supporting advances in object detection, semantic segmentation, scene understanding, and sensor robustness.
Beyond autonomous driving, the research has potential applications in intelligent transportation systems, advanced driver assistance systems (ADAS), smart cities, robotics, edge AI, and safety-critical artificial intelligence. It also provides an excellent educational resource for students and researchers working in computer vision, machine learning, and dependable AI, fostering collaboration and innovation across the academic community.
Through this public release, the IPTC researchers contribute to the development of safer, more reliable, and trustworthy AI systems capable of operating under diverse real-world environmental conditions.
A preview with sample images at https://ging.github.io/Weather-nuScenes/
You can access the dataset here: https://zenodo.org/records/21135951
Juan Diego Molero García – Morato: GS /LinkedIn
Javier Conde Díaz: GS / ORCID / LinkedIn
Gonzalo Martínez Ruiz de Arcaute: ORCID / LinkedIn
Carlos Arriaga Prieto: GS / ORCID / LinkedIn
Pedro Reviriego: GS / ORCID / LinkedIn
LinkedIn: https://www.linkedin.com/company/iptc-upm/
For more information: www.iptc.upm.es
Share this:
