Berlin-based automated driving data handling platform SiaSearch has partnered with Motional, a driverless technology specialist. The collaboration will provide enhanced access to nuScenes data sets for the self-driving research community, enabling users to access, explore and understand the data in greater depth than before.
SiaSearch is a platform which automatically extracts granular metadata from raw multimodal automated driving sensor data. It then makes the data fully searchable and accessible for evaluation through its GUI and API. With a catalog of over 50 different driving events and attributes, the company states users can now ask any questions they have about the raw data and get answers within seconds thanks to the proprietary Sia Interval Query Engine.
nuScenes is a dataset that teaches autonomous vehicles how to safely engage with ever-changing road environments. It is a diverse data set, comprising a full sensor suite with data from two distinct cities – Boston, Massachusetts, and Singapore – and manual annotations for 23 object classes. SiaSearch notes that finding edge cases and interesting interactions within a large scale data set, such as nuScenes, is like finding a needle in a haystack.
By making nuScenes available through SiaSearch, the data set is even more powerful, as it becomes fully accessible and the process of selecting the relevant driving sequences is streamlined.
For example, the system automatically tags nuScenes with driving maneuvers, interactions with other traffic agents, and infrastructural and environmental attributes. With the tagging, researchers can use nuScenes to find any driving scenario they need within seconds, enabling them to focus on their job’s essence, training models, instead of spending precious time on manually reviewing the data set.
The enhancements to the nuScenes data set can be accessed through the SiaSearch website. Engineers, researchers and other industry experts can register now and will receive free access, for non-commercial use, following a brief approval process.