why "xyz_cam0 = utils.geom.apply_4x4(cams_T_velo[:,0], xyz_velo0)"?

[yu-Mas]

Thanks for your great work. When i read the code, i meet a question in the line 145 of "train_nuscenes.py". xyz_cam0 = utils.geom.apply_4x4(cams_T_velo[:,0], xyz_velo0) "cams_T_velo[:,0]" means the "camera0 to world" and "xyz_velo0" means the points in the world coordinate.

It seems that followed code is right. xyz_cam0 = utils.geom.apply_4x4(velo_T_cams[:,0], xyz_velo0)

Except your reply. Thanks.

[yu-Mas]

And there is another question: intrinsics mean camera coordinate to image pixel coordinate,but variable "pix_T_cams_" seems represent from image pixel coordinate to camera coordinate.It confuses me for some time. pix_T_cams_ = utils.geom.merge_intrinsics(*utils.geom.split_intrinsics(intrins_)).to(device) Thanks for your reply.

[aharley]

Thanks for asking. It looks like you may be interpreting the notation incorrectly. If you have a look at the "Geometry conventions" stated in the readme, these variables and matmuls should make more sense.

[yu-Mas]

  您好，邮件已收到，我会尽快处理，谢谢。

[wcyjerry]

Hi, I also find this hard to understanding, lrtlist_cam = utils.geom.apply_4x4_to_lrt(cam_T_velo[0:1].repeat(N_, 1, 1), lrtlist_).unsqueeze(0), and ChatGPT 4.0 seems fail to explain this neither. as lrtlist_ is already under self-car coordinates, shouldn't use velo_T_cam，why cam_T_velo

[wcyjerry]

@yu-Mas Have you figured this out?

[yu-Mas]

  您好，邮件已收到，我会尽快处理，谢谢。