The study introduces a semi-automatic method for generating ground-truth estimation with the use of LiDAR sensors on vehicles and roadside units. The data from these sensors is converted into time-stamped observation grids, which are then fused to create dynamic reference information. The method is validated through several datasets. The researchers also released a dataset featuring real data from three vehicles and a roadside sensor. This research provides a new approach to evaluating the performance of perception systems in intelligent vehicles.
Publication date: 2023-07-24
Project Page: https://hal.science/hal-04172876
Paper: https://arxiv.org/pdf/2311.10415
