The aim of SafeSign is to increase confidence in driver assistance systems.
The increasing use of automated systems on the road (advanced driver assistance systems, autonomous vehicles) requires that traffic signs can be perceived with high quality by these systems. Especially during the transition from human to fully automated systems (mixed usage of road infrastructure) a different interpretation of signs exposes high risks for road safety.
In this project we investigate the influence of disturbance on deep learning based traffic sign classification systems. Humans typically can not judge if a sign can perceived correctly by an Artificial Intelligence. For example: a human driver approaches a tunnel. The variable traffic sign (based on LEDs) clearly shows “STOP” due to a height incident in the tunnel. The human driver stops the car. Due to disturbances (e.g., weather condition such as snow, ice, fog; defect of some LEDs) the automated system of the following car does not (or not correctly) perceive the sign. A crash happens.
In order to provide trust in future mixed usage situations, we combine and develop methods to improve deep learning robustness and interpretability. We evaluate the methods on real images of traffic signs (with / without disturbances) as well as synthetically generated disturbance images. The technical methods and concepts are constantly reviewed with respect to ethical principles.
The main project results (e.g., prototypical deep learning models, a database of disturbed traffic sign images) will be made available for the public. Austrian companies in the field of mobility, road infrastructure and autonomous driving can start further research and development project based on this results. In particular, we expect guidance for the design of future (human and machine readable) traffic signs.
Research that is based on ethical principles and guidelines will increase trust in Artificial Intelligence systems and boost social acceptance of such systems. In particular for safety relevant areas, such as traffic, this is of utmost importance.
AI-based pipeline for traffic sign detection and classification
Development of ATSD. 7555 frames containing traffic signs were extracted from HD videos and then annotated. After filtering invalid images, the remaining 7454 frames were then split into training, test and internal sets, forming ATSD-Scenes. ATSD-Signs consists of traffic sign patches extracted from the frames in the respective splits.
Some example scenes of ATSD, highlighting the diversity of the data set. It covers rural, urban and mountainous areas, and lots of tunnels.
This project is funded by the FFG.
Project Manager & Senior Researcher
Current vehicles use artificial intelligence (AI) to recognize traffic signs in order to inform drivers or adjust the speed of the vehicle. We often blindly trust these systems — but what are the limits of machine perception? In Crash Me If You Can we playfully get to the bottom of this question. Visitors have the opportunity to manipulate traffic signs in such a way that they are no longer correctly recognized by the AI. If a speed limit on the miniature racetrack is recognized incorrectly, the racing car flies out of the curve. RISC Software GmbH is dealing with these and other issues relating to traffic sign recognition by AI in the ”SafeSign” research project. This is being funded by the Austrian Research Promotion Agency (FFG) as part of the Ideas Lab 4.0 program. The AI research is supported by funds from the strategic economic and research program ”Innovative Upper Austria 2020” organized by the state of Upper Austria.
The exhibit was already on display at the Ars Electronica Festival 2021 and the Long Night of Research 2022.
