I do not know whether you have considered this, but in monocular VO with bundle adjustment it may be interesting to include the scale directly in the similarity transformation (see http://roboticsproceedings.org/rss06/p10.pdf, page 6) and optimize over it rather than separately.

HI
thanks for your release.
recently when i use liegroups, here appears an strange error.
AttributeError: Can't get attribute 'SO3' on <module 'liegroups.torch.so3' from 'PATH'>
do you have any idea about this problem?

Code to reproduce

R = np.array([[-0.26588199,  0.0721791,  -0.96129961],
                      [ 0.01772801,  0.99739046,  0.06998566],
                      [ 0.96384256,  0.001566,   -0.26646775]])
R = SO3.from_matrix(R, True)
q = R.to_quaternion()
print(np.isclose(la.norm(q), 1))

prints False

Hi, thank you for professional Liegroup package.

I can get different results when trying left multiply disturbance by two ways :

T : a 4x4 transition numpy array
delta_t=np.array([1,0,0,0,0,0]) #liealgebra

print(SE3.exp(delta_t).as_matrix().dot(T)) #1st way
print(SE3.exp(delta_t).dot(T)) #2nd way which delivered a wrong result

Should the result from the 2nd way be equal to the 1st way from your original idea?
Or, is T an illegal input in the 2nd way?

As the tilte, thanks ahead!

Hello,

In the so3.log function

There's a division by np.sin(angle) but there isn't a check to see if angle is +np.pi or -np.pi. I understand that due to numerical precision, np.sin(np.pi) != 0.0 but I'm wondering if there needs to be some extra care taken here.

See

https://en.wikipedia.org/wiki/Axis–angle_representation#Log_map_from_SO(3)_to_%7F'%22%60UNIQ--postMath-0000000D-QINU%60%22'%7F(3)

When I use liegroups in a loss function, it reports error as follows:
Can't call numpy() on Variable that requires grad. Use var.detach().numpy() instead.

Here is the code:

    print('rotation_ab_pred size = {} type = {}'.format(rotation_ab_pred.size(), rotation_ab_pred.type()))
    rvec_ab_pred = SO2.from_matrix(rotation_ab_pred)

And the result:

rotation_ab_pred size = torch.Size([32, 2, 2]) type = torch.cuda.FloatTensor

Hi,
first of all thanks for your work.
My problem is linked to transformation of SE3 matrix to xi vector and vice versa. This code section below is a test that shows that the original SE3 matrix has different values compare to output matrix, but this happens with just values that have small module.

import torch
from liegroups.torch import SE3

print("Test 1:")
test_SE3_matrix = torch.tensor([[0.9999957085, -0.002008878859, -0.002134637441, 1.187580466], [0.002002746798, 0.9999938607, -0.002870822791, 0.06579889357], [0.002140391385, 0.002866535215, 0.9999936223, -0.05322432518], [0., 0., 0., 1.]])
print("Testing SE3 matrix:\n", test_SE3_matrix)
test_xi = SE3.from_matrix(test_SE3_matrix).log()
print("Output xi value:", test_xi)
test_original_value_SE3_matrix = SE3.exp(test_xi).as_matrix()
print("It should be the same as the start test matrix:\n", test_original_value_SE3_matrix)

print("\nTest 2:")
test_SE3_matrix_2 = torch.tensor([[1.000000e-00, 9.043681e-12, 2.326810e-11, -5.551115e-17], [9.043682e-12, 1.000000e+00, 2.392370e-10, 0.000000e+00], [2.326809e-11, 2.392370e-10, 9.999999e-01, -2.220446e-16], [0., 0., 0., 1.]])
print("Testing SE3 matrix:\n", test_SE3_matrix_2)
test_xi_2 = SE3.from_matrix(test_SE3_matrix_2).log()
print("Output xi value:", test_xi_2)
test_original_value_SE3_matrix_2 = SE3.exp(test_xi_2).as_matrix()
print("It should be the same as the start test matrix:\n", test_original_value_SE3_matrix_2)

And this is the output:

Test 1:
Testing SE3 matrix:
tensor([[ 1.0000, -0.0020, -0.0021, 1.1876], [ 0.0020, 1.0000, -0.0029, 0.0658], [ 0.0021, 0.0029, 1.0000, -0.0532], [ 0.0000, 0.0000, 0.0000, 1.0000]])
Output xi value: tensor([ 1.1876, 0.0645, -0.0546, 0.0029, -0.0021, 0.0020])
It should be the same as the start test matrix:
tensor([[ 1.0000, -0.0020, -0.0021, 1.1876], [ 0.0020, 1.0000, -0.0029, 0.0658], [ 0.0021, 0.0029, 1.0000, -0.0532], [ 0.0000, 0.0000, 0.0000, 1.0000]])

Test 2:
Testing SE3 matrix:
tensor([[ 1.0000e+00, 9.0437e-12, 2.3268e-11, -5.5511e-17], [ 9.0437e-12, 1.0000e+00, 2.3924e-10, 0.0000e+00], [ 2.3268e-11, 2.3924e-10, 1.0000e+00, -2.2204e-16], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, 1.0000e+00]])
Output xi value: tensor([-5.5511e-17, -2.6310e-26, -2.2204e-16, 2.3924e-10, 2.3268e-11, 9.0437e-12])
It should be the same as the start test matrix:
tensor([[ 1.0000e+00, -9.0437e-12, 2.3268e-11, -5.5511e-17], [ 9.0437e-12, 1.0000e+00, -2.3924e-10, 0.0000e+00], [-2.3268e-11, 2.3924e-10, 1.0000e+00, -2.2204e-16], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, 1.0000e+00]])

As you can see, in Test 2, the element (3,1), (1,2) and (2,3) have wrong sign.

I recently found two occurrences of this error, both in torch/so3.py:

Traceback (most recent call last):
File "", line 1, in
File "/home/ubuntu/.local/lib/python3.6/site-packages/liegroups/torch/se3.py", line 80, in exp
rot_jac = cls.RotationType.left_jacobian(phi)
File "/home/ubuntu/.local/lib/python3.6/site-packages/liegroups/torch/so3.py", line 187, in left_jacobian
#large_angle_mask = 1 - small_angle_mask # element-wise not
File "/home/ubuntu/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/torch/tensor.py", line 362, in rsub
return _C._VariableFunctions.rsub(self, other)
RuntimeError: Subtraction, the - operator, with a bool tensor is not supported. If you are trying to invert a mask, use the ~ or logical_not() operator instead.

I could verify locally that it is just a matter of doing what is suggested and it works fine.
P.S.: My pytorch version is 1.3

This assertion fails.

import numpy as np
from liegroups import SO3
s = SO3.from_rpy(np.pi, 0, np.pi/2)
es = SO3.exp(s.log())
assert np.allclose(s.mat, es.mat)