scene flow estimation is the task of describing the 3D motion field between
temporally successive point clouds. State-of-the-art methods use strong priors
and test-time optimization techniques, but require on the order of tens of
seconds for large-scale point clouds, making them unusab
本文提出了一种基于生成的深度图的方法,通过引入像素的稠密性来实现直接从 2D 图像学习 3D 场景流,以及利用统计方法和视差一致性损失来解决噪声点的问题,从而达到了更加有效的自监督学习 3D 场景流的目的。实验证明,这种方法优于合成数据集和激光雷达点云学习的方法,在场景流估计任务中表现出更好的稳定性和准确度。