questions about transformations about persformer_3dlane HOT 8 CLOSED

0. the transformation is largely following GenLaneNet. You can have a try on R_gv@E method and see if there's any improvement.
1. sorry, that was a typo in OpenLane repo. We'll fix it. The correct coordinate system is vehicle coordinates.

from persformer_3dlane.

for this step

 cam_extrinsics[:3, :3] = np.matmul(np.matmul(np.matmul(np.linalg.inv(R_vg),cam_extrinsic[:3, :3]),R_vg), R_gc)

could you help me figure out, what is this inv(R_vg) @ E @ R_vg @ R_gc doing? To my mind, E_g2c is final target here, am I right? But I still can not understand each step transformation in above line. Could you give some interpretation that would be very appreciated :) thanks !

from persformer_3dlane.

same question as above

from persformer_3dlane.

@ChonghaoSima it seems quite strange that the annotations are in the vehicle_coordinate system. Specifically, the z part of the annotations tends to be negative most of the times. it seems to in the camera coordinates after all.
For example, in lane3d_1000/training/segment-15832924468527961_1564_160_1584_160_with_camera_labels/150767882737686600.json the first lane is (labels_dict is dictionary of the loaded json)

min(lanes_visualization_obj.labels_dict['lane_lines'][0]['xyz'][2])
Out[15]: -3.8558186921526554
max(lanes_visualization_obj.labels_dict['lane_lines'][0]['xyz'][2])
Out[16]: -2.2029994863655915

And the 2nd one is:

 min(lanes_visualization_obj.labels_dict['lane_lines'][1]['xyz'][2])
Out[18]: -2.4630904216158926
max(lanes_visualization_obj.labels_dict['lane_lines'][0]['xyz'][2])
Out[19]: -2.2029994863655915

and the third one is:

min(lanes_visualization_obj.labels_dict['lane_lines'][2]['xyz'][2])
Out[20]: -3.8966073247935045
max(lanes_visualization_obj.labels_dict['lane_lines'][2]['xyz'][2])
Out[21]: -2.4456845933203346

from persformer_3dlane.

Please refer to https://github.com/OpenPerceptionX/OpenLane/issues/24. It should solve your questions. The annotations are under the Waymo camera coordinate (The typo has been updated in https://github.com/OpenPerceptionX/OpenLane/blob/main/anno_criterion/Lane/README.md).

from persformer_3dlane.

We will close this issue, if you still have the same questions, please reopen this.

from persformer_3dlane.

It should solve your questions. The annotations are under the Waymo camera coordinate (The typo has been updat

Seems the link is invalid, may you update the explanation of this issue?

from persformer_3dlane.

@usherbob Just change OpenPerceptionX to OpenDriveLab: https://github.com/OpenDriveLab/OpenLane/blob/main/anno_criterion/Lane/README.md

from persformer_3dlane.