[robotics-worldwide] [news] Announcing nuScenes - the largest, most complex, and diverse self-driving cars perception dataset

Andrea Censi andrea at nutonomy.com
Fri Mar 29 10:36:23 PDT 2019


Dear colleagues,

We at nuTonomy / Aptiv believe that the fastest way to obtain
super-human safety in self-driving cars is for industry to have a
collaborative, open, and transparent approach.

Today, we are excited to announce the 1.0 release of the nuScenes dataset.

nuScenes is our contribution to the field of perception. nuScenes is
the largest, richest, and most diverse dataset for self-driving cars
perception. This is an industry-grade dataset in scale and complexity,
and representative of the datasets we use internally for development.

nuScenes is a multimodal dataset. It contains data from a full sensor
suite of a modern self-driving car: 1 LIDAR, 5 RADARs, 6 cameras, IMU
and GPS. The dataset has 1.4 million images, 390K LiDAR sweeps, and
1.4M 3D human annotated bounding boxes.

The data is collected in Boston and Singapore (left and right driving,
hot and cold), and sampled in the most complex driving scenarios in
each urban environment.

You can find all details at the website: https://urldefense.proofpoint.com/v2/url?u=https-3A__www.nuScenes.org&d=DwIBaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=m_2D6Cw5xAKfyOBFjLNIZLUxBk4eK7q3NOoLu0ORuTw&s=KJyIw0G1X5BPSEiKS0dQ4UUqyzHzoM3yhPbY7-HbW5A&e=

We are hosting the first nuScenes detection challenge at CVPR 2019. We
are looking forward to seeing your submissions!
https://urldefense.proofpoint.com/v2/url?u=https-3A__www.nuscenes.org_object-2Ddetection&d=DwIBaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=m_2D6Cw5xAKfyOBFjLNIZLUxBk4eK7q3NOoLu0ORuTw&s=kLAmdz7NTMlCDqedAUG28VCcthCpHo7tMH30wICmi5U&e=

We hope that your detectors and classifiers like our data.
Happy perception!

--
Andrea Censi
Director of Research, nuTonomy / Aptiv Autonomous Mobility


More information about the robotics-worldwide mailing list