Sensor models used for the recordings
Hi,
Thanks a lot for the amazing dataset. I have a question: Which sensor models have been used for the data capture? Also, are the data already kitti compatible? (I am asking for the lidar object detection using openPCDet, which requires the point cloud to be oriented with the KITTI coordinate system.)
Cheers
Thanks for your interest! Unfortunately we are not at liberty to state the exact sensor model numbers here (though I can say that this is all data captured from NVIDIA DRIVE Hyperion 8/8.1 platforms for which some public documentation/articles exist) but we have at least tried to indicate when sensors are different, e.g., this dataset includes data from 4 distinct radar sensors -- srr_0, lrr_1, mrr_2, and srr_3.
Regarding data format, we are in the process of preparing a developer kit (coming soon at https://github.com/NVlabs/physical_ai_av) to make data loading and format transformation more transparent -- getting this much data onto HF has required some rather aggressive packing. One thing I'll note for now, however, is that the LiDAR point clouds as decoded directly from serialized Draco are not egomotion compensated; we plan to include a utility to perform this motion compensation in the upcoming dev kit.