waymo数据集 Waymo无人驾驶高分辨率传感器数据集

Waymo 开放数据集由 Waymo 自动驾驶汽车在各种条件下收集的高分辨率传感器数据组成，它与 KITTI、NuScenes 等数据集的对比数据如下，在传感器配置、数据集大小上都有很大的提升。 Waymo 数据集的传感器包含 5 个激光雷达、5 个摄像头，激光雷达和摄像头的同步效果也更好。 更重要的是，Waymo 数据集包含 3000 段驾驶记录，时长共 16.7 小时，平均每段长度约为 20 秒。整个数据集一共包含 60 万帧，共有大约 2500 万 3D 边界框、2200 万 2D 边界框。 此外，在数据集多样性上，Waymo Open Dataset 也有很大的提升，该数据集涵盖不同的天气条件，白天、夜晚不同的时间段，市中心、郊区不同地点，行人、自行车等不同道路对象，等等。 1. 规模和覆盖范围： 数据集包含3000个驾驶片段，每一片段包含20秒的连续驾驶画面。连续镜头内容可以使得研究人员开发模型来跟踪和预测其他道路使用者的行为。 2. 多样化的驾驶环境： 数据采集的范围涵盖凤凰城、柯克兰、山景城、旧金山等地区，以及各种驾驶条件下的数据，包括白天、黑夜、黎明、黄昏、雨天和晴天。 3. 高分别率的特点和360度的视图： 每个分段涵盖5个高分辨率Waymo激光雷达和五个前置和侧面摄像头的数据。 4. 密集的标签信息： 车辆、行人、自行车、标识牌等图像都经过精心标记，一共打了2500万个3D标签和2200万个2D标签。 5. 是相机-激光雷达同步： Waymo称，该团队致力于融合了多个摄像头和激光雷达数据的3D感知模型。Waymo设计了全套的自动驾驶系统，包含硬件和软件，用以无缝地协同工作，其中包括选择传感器的位置和高质量的时间同步。 数据使用引用：

The Waymo Open Dataset consists of high-resolution sensor data collected by Waymo autonomous vehicles under various conditions. Compared with datasets such as KITTI and NuScenes, it offers significant improvements in sensor configuration and dataset scale.

The sensors of the Waymo dataset include 5 LiDARs and 5 cameras, with better synchronization between LiDARs and cameras. More importantly, the Waymo Open Dataset contains 3000 driving segments totaling 16.7 hours, with an average duration of approximately 20 seconds per segment. The entire dataset includes 600,000 frames, approximately 25 million 3D bounding boxes and 22 million 2D bounding boxes.

In addition, the Waymo Open Dataset also features substantial improvements in dataset diversity, covering different weather conditions, different time periods such as daytime and nighttime, various locations including downtown and suburban areas, and different road objects such as pedestrians and bicycles, among others.

1. Scale and Coverage: The dataset contains 3000 driving segments, each with 20 seconds of continuous driving footage. The continuous footage enables researchers to develop models for tracking and predicting the behaviors of other road users.
2. Diverse Driving Environments: Data was collected in areas including Phoenix, Kirkland, Mountain View, San Francisco, etc., under various driving conditions, including daytime, nighttime, dawn, dusk, rainy and sunny weather.
3. High-Resolution Features and 360-Degree Field of View: Each segment covers data from 5 high-resolution Waymo LiDARs and five front and side cameras.
4. Dense Labeling Information: Images of vehicles, pedestrians, bicycles, traffic signs and other objects have been meticulously annotated, with a total of 25 million 3D labels and 22 million 2D labels.
5. Camera-LiDAR Synchronization: Waymo states that its team is committed to developing 3D perception models that fuse data from multiple cameras and LiDARs. Waymo has designed a complete autonomous driving system, including hardware and software, to work seamlessly together, including the selection of sensor positions and high-quality time synchronization.

Data Usage Citation:

搜集汇总

背景与挑战

背景概述

waymo数据集是由Waymo公司提供的无人驾驶高分辨率传感器数据集，包含3000段驾驶记录，时长共16.7小时，涵盖多种驾驶条件和环境。数据集特点包括高分辨率和360度视图、密集的标签信息（2500万3D标签和2200万2D标签）以及相机-激光雷达同步，适用于自动驾驶相关研究。