naver/roma

RoMa: A lightweight library to deal with 3D rotations in PyTorch.

What it solves

It simplifies the complex mathematical process of handling 3D rotations in PyTorch, providing a way to convert between different rotation representations and perform operations on them in a way that is differentiable, which is essential for training machine learning models.

How it works

RoMa provides a set of tools to map between various rotation formats (such as rotation vectors, unit quaternions, and rotation matrices) and maps Euclidean space to rotation space. It includes utilities for calculating geodesic distances, performing spherical interpolation (Slerp), and managing rigid transformations (rotation and translation) using homogeneous matrices.

Who it’s for

Researchers and developers working on 3D vision, robotics, and machine learning projects that require precise and differentiable 3D rotation manipulation.

Highlights

  • Differentiable mappings between rotation representations.
  • Support for arbitrary numbers of batch dimensions.
  • Tools for rotation regression from arbitrary inputs using Special Procrustes and Special Gram-Schmidt orthonormalization.
  • Built-in support for rigid transformations and homogeneous matrices.

Related

  • Project
    facebookresearch/pytorch3d

    A PyTorch-based library for 3D Computer Vision research providing differentiable rendering and efficient tools for manipulating 3D meshes and point clouds.

  • Project
    sea-bass/pyroboplan

    An educational Python library for manipulator motion planning that uses Pinocchio for robot kinematics and dynamics.

  • Project
    lucidrains/ema-pytorch

    A PyTorch library for maintaining Exponential Moving Average (EMA) versions of models to improve stability and performance, featuring post-hoc synthesis and target representation routing.

  • Project
    facebookresearch/multimodal

    A PyTorch library for training state-of-the-art multimodal models at scale, providing modular building blocks and pretrained weights for models like CLIP and BLIP-2.

  • Project
    pypose/bae

    A PyTorch-based library for exact 2nd-order sparse optimization in robotics, specifically designed to accelerate Bundle Adjustment and Pose Graph Optimization using CUDA.