New lidar dataset captures motion for autonomous driving

Conventional lidar for self-driving cars measures where objects are; 4D FMCW lidar also measures how fast they're moving. This dataset from Beijing includes velocity readings from multiple lidar types plus cameras and ground truth, collected in dense urban scenes with pedestrians and high-speed traffic. Velocity-aware detection and motion forecasting improve especially for vulnerable road users and fast objects—suggesting this sensing mode is worth investing in. Dataset and benchmarks are open-sourced.