The origin code in argoverse_map_tutorial shows the prejection result of the centerlines on image. This step includes filtered the centerlines which is occluded. The function "plot_lane_centerlines_in_img2" provides this. But when filtering, the "clip_point_cloud_to_visible_region" is operating on the unsync lidar(ego) frame and camera frame. To correct this:
Origin:
centerline_egovehicle_fr = city_SE3_egovehicle.inverse().transform_point_cloud(centerline_city_fr)
centerline_uv_cam = cam_SE3_egovehicle.transform_point_cloud(centerline_egovehicle_fr)
can also clip point cloud to nearest LiDAR point depth
The origin code in argoverse_map_tutorial shows the prejection result of the centerlines on image. This step includes filtered the centerlines which is occluded. The function "plot_lane_centerlines_in_img2" provides this. But when filtering, the "clip_point_cloud_to_visible_region" is operating on the unsync lidar(ego) frame and camera frame. To correct this: Origin: centerline_egovehicle_fr = city_SE3_egovehicle.inverse().transform_point_cloud(centerline_city_fr) centerline_uv_cam = cam_SE3_egovehicle.transform_point_cloud(centerline_egovehicle_fr)
can also clip point cloud to nearest LiDAR point depth
Repair: centerline_egovehicle_fr = city_SE3_egovehicle.inverse().transform_point_cloud(centerline_city_fr)
can also clip point cloud to nearest LiDAR point depth
This visualization image result shows correctly.