nuScenes by Aptiv: Leading Trust and Collaboration in the AV Industry

Posted By
Oscar Beijbom
Machine Learning Director, Aptiv

As the first to publicly launch our open source data set in March 2019, we are excited to see the recent announcement by our partner Lyft, joining us in the release of their open source data set. It is also exciting to see other companies planning to follow suit with their own releases: Waymo, Hesai, Argo and Zoox.

Historically, competition between AV developers and operators meant that data and information from years of testing had been kept extremely private. This ultimately resulted in a lack of industry-wide collaboration and knowledge sharing in the development of safe AV technology. Providing public data not only has the potential to support academic researchers and industry experts’ access to carefully curated safety standards. It also enables robust progress and innovation in the industry.

With the March 2019 release of nuScenes, our team led an important change in the long-held industry habit of keeping information secret. As the first large-scale public dataset to provide information from a comprehensive AV sensor suite, nuScenes is built on a mission to enable future mobility through transparency and trust.

A diverse set of data collected from Boston and Singapore, nuScenes is organized into 1,000 “scenes” and is representative of some of the most complex driving scenarios in each urban environment. The dataset is composed of 1.4 million images, 390K lidar sweeps, and 1.4M 3D human annotated bounding boxes, representing the largest multimodal 3D AV dataset at the time of release.

Since the launch in March, over 3,500 users have registered from institutions like MIT, Harvard, CMU, Stanford, Berkeley, and ETH Zurich; as well as industry players such as Google, Apple, Amazon, Facebook, Microsoft, NVIDIA, Uber, Toyota, Ford, Honda, Alibaba, ASUS, NEC, IBM, Samsung, Xiaomi, Renault, Yandex. Most recently, the Aptiv team co-organized CVPR 2019’s Workshop on Autonomous Driving where we hosted the first  nuScenes detection challenge.

We continue to believe there is a critical opportunity to do more as an industry to collaborate on safety. In addition to the nuScenes release in March, we also joined industry leaders across the automotive and automated driving technology spectrum in July and published “Safety First for Automated Driving,” (SaFAD), a non-binding organized framework for the development, testing and validation of safe automated passenger vehicles.

These 11 leaders — Aptiv, Audi, Baidu, BMW, Continental, Daimler, Fiat Chrysler Automobiles, HERE, Infineon, Intel, and Volkswagen — comprise the broadest representation across the industry and have published, to date, the largest report on how to build, test and operate a safe automated vehicle. The SaFAD white paper authors’ purpose is to emphasize the importance of safety by design, along with verification and validation, as the industry works toward creating standards for automated driving.

At Aptiv, we believe that safety transcends competition, and look forward to continued collaboration across the industry. To explore our dataset, visit