Skip to main content

MobilityXlab startups and Zenseact collaborate in extensive dataset to advance AD technology

Monday, April 15, 2024

Fulfilling the potential of Autonomous Driving (AD) to reach zero deaths caused by traffic accidents requires collaboration and openness. Zenseact, one of MobilityXlab’s partners, embraced this principle with the release of the Zenseact Open Dataset (ZOD), a large multi-modal AD dataset created by its researchers with contributions from two MobilityXlab startups: brighter AI and Kognic.

A photo showing the zenseact logo on their office building.

Photo: Lindholmen Science Park

The reliability and safety of autonomous driving systems rely directly on the quality of the data their models are trained upon. Diverse and comprehensive datasets ensure that the system is exposed to a wide range of scenarios, including rare or unexpected ones, enabling it to respond appropriately in different situations. To ensure that, ZOD contains data collected over a 2-year period in 14 different European counties, using a fleet of vehicles equipped with a full sensor suite. 

According to Mina Alibeigi, AI Researcher and ZOD team project leader at Zenseact, making it available to the entire community was essential. She says: We want to encourage the whole AD community to improve the advancement of its systems. Because we, at Zenseact, have this zero vision in mind, it means that if we can focus and work on these questions together, then we can solve the challenges faster.

The dataset is currently the only one released under the permissive CC BY-SA 4.0 license, allowing for both commercial and non-commercial use for research, benchmarking, and development.

But making such a complex dataset widely available has also its challenges: the data needs to be ready to be used and it must be privacy-compliant. “So it's important that, for instance, when we want to share such data, we make sure that we protect the privacy of humans and other vehicles around us. So we needed to anonymize, for instance, the license plates and human faces,” explains Mina Alibeigi

And here collaboration was also key. Brighter AI, one of the startups in MobilityXlab’s portfolio, provided Deep Neural Anonymization Technology (DNAT) and blurring to protect all personally identifiable information in the images, including faces and license plates. 

The relationship between the two companies started in 2019, when Brighter AI was selected to join MobilityXlab’s Batch 4 and subsequently developed a proof of concept (PoC) with Zenseact. 

Kognic, one of the first startups that joined MobilityXlab has had also an important role. By providing high-quality and detailed annotation, it allowed for the creation of a development toolkit, making the dataset easy to get started with. 

The Zenseact Open Dataset is a clear example of the potential that can be unleashed with collaboration between startups and companies: high-quality, privacy-safe, ready-to-use data that can advance autonomous driving and get society closer to Zero Vision.

To learn more about the Zenseact Open dataset and get access to it, visit: https://zod.zenseact.com/