One Thousand and One Hours: Self-driving Motion Prediction Dataset

By J. Houston, Guido C. A. Zuidhof, Luca Bergamini, Yawei Ye, Ashesh Jain, Sammy Omari, Vladimir I. Iglovikov, Peter Ondruska. Conference on Robot Learning, 2020

483 citations recorded by the source corpus.

We present the largest self-driving dataset for motion prediction to date, with over 1,000 hours of data. This was collected by a fleet of 20 autonomous vehicles along a fixed route in Palo Alto, California over a four-month period. It consists of 170,000 scenes, where each scene is 25 seconds long and captures the perception output of the self-driving system, which encodes the precise positions and motions of nearby vehicles, cyclists, and pedestrians over time. On top of this, the dataset contains a high-definition semantic map with 15,242 labelled elements and a high-definition aerial view over the area. Together with the provided software kit, this collection forms the largest, most complete and detailed dataset to date for the development of self-driving, machine learning tasks such as motion forecasting, planning and simulation. The full dataset is available at this http URL.

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/5ff9429efe7b6c4e7c039064920600e2b7664828

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Automotive Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting (2023, methodology): “Also influential are self-driving “motion prediction” datasets [11, 19, 29, 4, 45] — containing abstracted object tracks instead of raw sensor data — of which the Argoverse Motion Forecasting dataset [6] was the first.”
Explanations in Autonomous Driving: A Survey (2021, background): “This information can be difficult to assess from AVs at present but, as a starting point, driving simulators and the Lyft Level5 dataset are helpful to create a prototype at the least.”
nuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles (2021, methodology): “The Lyft Level 5 Prediction Dataset [11] contains 1118h of data from a single route of 6.”
Cross-view Transformers for real-time Map-view Semantic Segmentation (2022, background): “Driven by ever larger 3D recognition datasets [3, 5, 10, 13, 38], a number of works have focused on perception in the map-view.”
How many Observations are Enough? Knowledge Distillation for Trajectory Forecasting (2022, background): “[19] is one of the largest collection of traffic agent motion data.”
Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion Dataset (2021, background): “Recently released datasets for motion forecasting [9, 18, 4] are orders of magnitude larger than popular 3D perception datasets [4, 19, 32, 15].”
End-to-End Urban Driving by Imitating a Reinforcement Learning Coach (2021, background): “Off-policy data for AD includes, for example, public driving datasets [6, 22, 51].”
Wayformer: Motion Forecasting via Simple & Efficient Attention Networks (2022, background): “a) Modality specialized architectures: Successful modeling techniques for motion forecasting fuse multi-modal inputs that represent different static, dynamic, social and temporal aspects of the scene [9], [22], [13], [40], [8], [45].”