Argoverse 自动驾驶

Argoverse 数据集包含 3D Tracking 和 Motion Forecasting 两部分，Argoverse 3D tracking 数据集包含 113 个场景的 3d 跟踪注释。每个片段长度为 15-30 秒，共计包含 11319 个跟踪对象。在训练集和测试集的每个片段场景中包含了五米内的所有物体的注释，可被理解为检测汽车可驾驶区域（5 米）的所有物体，以 3d 框架形式展现。该数据集可被自动驾驶等领域使用。 Argoverse Motion Forecasting 数据集为运动预测类模型的数据集，包含 327793 个场景，每个场景时常 5 秒，且包含以 10 Hz 采样的每个跟踪对象的 2D 鸟瞰图。该数据集是由超过 1000 小时的街道驾驶所获取，可用于自动驾驶等领域的研究。

The Argoverse dataset includes two components: 3D Tracking and Motion Forecasting.
The Argoverse 3D Tracking dataset contains 3D tracking annotations for 113 scenarios. Each segment has a duration of 15 to 30 seconds, with a total of 11,319 tracked objects. Annotations for all objects within a 5-meter range are provided for every segment in both the training and test sets, which can be interpreted as annotating all objects in the vehicle's drivable area (within 5 meters) in a 3D framework. This dataset can be utilized in fields such as autonomous driving.
The Argoverse Motion Forecasting dataset is tailored for motion prediction models, comprising 327,793 scenarios. Each scenario lasts 5 seconds, and includes 2D bird's-eye view annotations of each tracked object sampled at 10 Hz. This dataset is collected from over 1,000 hours of street driving, and can be employed for research in autonomous driving and related fields.

搜集汇总

背景与挑战

背景概述

Argoverse自动驾驶数据集包含3D Tracking和Motion Forecasting两部分，用于自动驾驶研究。3D Tracking部分涵盖113个场景的3D跟踪注释，包含11319个跟踪对象；Motion Forecasting部分包含327793个场景的2D鸟瞰图数据。该数据集总大小为260.38 GB，发布于7年前，基于超过1000小时的街道驾驶数据采集。