about can_bus?

[zhjw0927]

hello, What is the meaning of the variable 'can_bus' name?

In actual operation, the following phenomena are observed. After the four properties of the quaternion are assigned to the array, only the first property value is actually copied:

can_bus = np.zeros(18)
rotation = Quaternion._from_matrix(np.array(info['pose']['rotation']))
can_bus[:3] = info['pose']['translation']
can_bus[3:7] = rotation 

print(rotation) ------> Quaternion(0.7405030732851444, 0.004200992255077698, 0.01887259332437053, -0.6717747951066204)

print(can_bus) -------> array([6.65756166e+03, 1.83486642e+03, 5.96497427e+01, 7.40503073e-01,
       7.40503073e-01, 7.40503073e-01, 7.40503073e-01, 0.00000000e+00,
       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
       0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
       0.00000000e+00, 0.00000000e+00])

Is this expected? or Is it pyquaternion library version issue? Thank you!

[sephyli]

Good catch! It's not expected. But fortunately it is not used in the current model.在 2024年10月11日，17:15，zhjw0927 ***@***.***> 写道： hello, What is the meaning of the variable 'can_bus' name? In actual operation, the following phenomena are observed. After the four properties of the quaternion are assigned to the array, only the first property value is actually copied: can_bus = np.zeros(18) rotation = Quaternion._from_matrix(np.array(info['pose']['rotation'])) can_bus[:3] = info['pose']['translation'] can_bus[3:7] = rotation print(rotation) ------> Quaternion(0.7405030732851444, 0.004200992255077698, 0.01887259332437053, -0.6717747951066204) print(can_bus) -------> array([6.65756166e+03, 1.83486642e+03, 5.96497427e+01, 7.40503073e-01, 7.40503073e-01, 7.40503073e-01, 7.40503073e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00]) Is this expected? Thank you! —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[zhjw0927]

hello, please one more question. OpenLaneV2 provides only the image domain gt of the traffic element. Is this label sufficient for the task of topo relationship prediction, or is there no way to make BEV position based on argoverse data？
Thank you!