RoMa (which stands for Rotation Manipulation) provides differentiable mappings between 3D rotation representations, mappings from Euclidean to rotation space, and various utilities related to rotations.
It is implemented in PyTorch and aims to be an easy-to-use and reasonably efficient toolbox for Machine Learning and gradient-based optimization.
Latest documentation is available here: https://naver.github.io/roma/.
Below are some examples of use of RoMa:
import torch
import roma
# Arbitrary numbers of batch dimensions are supported, for convenience.
batch_shape = (2, 3)
# Conversion between rotation representations
rotvec = torch.randn(batch_shape + (3,))
q = roma.rotvec_to_unitquat(rotvec)
R = roma.unitquat_to_rotmat(q)
Rbis = roma.rotvec_to_rotmat(rotvec)
# Regression of a rotation from an arbitrary input:
# Special Procrustes orthonormalization of a 3x3 matrix
R1 = roma.special_procrustes(torch.randn(batch_shape + (3, 3)))
# Conversion from a 6D representation
R2 = roma.special_gramschmidt(torch.randn(batch_shape + (3, 2)))
# From the 10 coefficients of a 4x4 symmetric matrix
q = roma.symmatrixvec_to_unitquat(torch.randn(batch_shape + (10,)))
# Metrics on the rotation space
R1, R2 = roma.random_rotmat(size=5), roma.random_rotmat(size=5)
theta = roma.utils.rotmat_geodesic_distance(R1, R2)
cos_theta = roma.utils.rotmat_cosine_angle(R1.transpose(-2, -1) @ R2)
# Operations on quaternions
q_identity = roma.quat_product(roma.quat_conjugation(q), q)
# Spherical interpolation between rotation vectors (shortest path)
rotvec0, rotvec1 = torch.randn(batch_shape + (3,)), torch.randn(batch_shape + (3,))
rotvec_interpolated = roma.rotvec_slerp(rotvec0, rotvec1, steps)
# Rigid transformation T composed of a rotation part R and a translation part t
t = torch.randn(batch_shape + (3,))
T = roma.Rigid(R, t)
# Composing and inverting transformations
identity = T @ T.inverse()
# Casting the result to a batch of 4x4 homogeneous matrices
M = identity.to_homogeneous()The easiest way to install RoMa is to use pip:
pip install roma
Alternatively one can install the latest version of RoMa directly from the source repository:
pip install git+https://github.com/naver/roma
For pytorch versions older than 1.8, we recommend installing torch-batch-svd
to achieve a significant speed-up with special_procrustes on CUDA GPUs.
You can check that this module is properly loaded using the function roma.utils.is_torch_batch_svd_available().
With recent pytorch installations (torch>=1.8), torch-batch-svd is no longer needed or used.
RoMa, Copyright (c) 2021 NAVER Corp., is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 license (see license).
Bits of code were adapted from SciPy. Documentation is generated, distributed and displayed with the support of Sphinx and other materials (see notice).
For a more in-depth discussion regarding differentiable mappings on the rotation space, please refer to:
Please cite this work in your publications:
@inproceedings{bregier2021deepregression,
title={Deep Regression on Manifolds: a {3D} Rotation Case Study},
author={Br{\'e}gier, Romain},
journal={2021 International Conference on 3D Vision (3DV)},
year={2021}
}