Generating optical flow tags from point clouds

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating optical flow tags from lidar point clouds. One of the methods includes obtaining data specifying a training sample, including a first image of a scene in an environment captured at a first point in time and a second image of the scene in the environment captured at a second point in time. For each of the plurality of lidar points, a respective second corresponding pixel in the second image is obtained, and a respective velocity estimate for the lidar point at a second point in time is obtained. A respective first corresponding pixel in the first image is determined using a velocity estimate of the lidar point. A proxy optical flow reference truth for the training sample is generated based on an estimate of optical flow for pixels between the first image and the second image. 用于从激光雷达点云生成光流标签的方法、系统和装置，包括在计算机存储介质上编码的计算机程序。方法之一包括获得指定训练样本的数据，包括在第一时间点捕获的环境中的场景的第一图像和在第二时间点捕获的环境中的场景的第二图像。对于多个激光雷达点中的每个，获得在