PoseLib/PoseLib

Minimal solvers for calibrated camera pose estimation

What it solves

PoseLib is a library of minimal solvers for camera pose estimation. It addresses the challenge of determining a camera's position and orientation (pose) based on different types of correspondences, such as point-to-point, point-to-line, and line-to-line matches between 2D images and 3D worlds or between two 2D images.

How it works

The library provides a collection of standalone, fast solvers that implement state-of-the-art algorithms for absolute and relative pose estimation. It uses a consistent calling interface and minimizes external dependencies (relying primarily on Eigen).

To handle noise and outliers in data, it implements robust estimators based on LO-RANSAC, which generate hypotheses using the minimal solvers and then perform non-linear refinement (minimizing reprojection or Sampson error) to find the most accurate pose.

It supports various camera models (compatible with COLMAP) and provides Python bindings for easy integration into other workflows.

Who it’s for

This library is designed for developers and researchers working in computer vision, SLAM (Simultaneous Localization and Mapping), and 3D reconstruction who need high-performance, robust tools for camera pose estimation.

Highlights

  • Diverse Solver Support: Includes solvers for absolute pose, relative pose, homography, and fundamental matrices.
  • Robust Estimation: Built-in LO-RANSAC and non-linear refinement (Bundle Adjustment) with multiple loss functions (Huber, Cauchy, etc.).
  • Monocular Depth Integration: Specialized solvers that utilize neural network-based monocular depth estimates to improve relative pose estimation.
  • Minimal Dependencies: Designed to be lightweight and easy to extract specific solvers for integration into other frameworks.
  • COLMAP Compatibility: Uses camera models and data formats compatible with COLMAP.

Related

  • Project
    nianticlabs/map-free-reloc

    A reference implementation for map-free visual relocalization that enables metric pose estimation relative to a single image, removing the need for large 3D maps.

  • Project
    perfanalytics/pose2sim

    A markerless 3D kinematics workflow that converts 2D video from low-cost cameras into research-grade 3D joint angles and skeletal analysis using OpenSim.

  • Project
    MIT-SPARK/KISS-Matcher

    A fast and robust point cloud registration library that aligns 3D point clouds to determine relative rotation and translation.

  • Project
    ACFR-RPG/DynOSAM

    DynoSAM is a Stereo/RGB-D Visual Odometry pipeline for Dynamic SLAM that estimates camera poses and object motions while mapping static backgrounds.

  • Project
    Tau-J/rtmlib

    rtmlib is a lightweight Python library for human and animal pose estimation that wraps RTMPose and ViTPose models. It avoids heavy OpenMMLab dependencies, requiring only numpy, OpenCV and ONNX Runtime (with optional GPU/accelerator back‑ends). The package provides high‑level solution classes (Wholebody, Body, Hand, Animal, Custom, Wholebody3d) that combine a detector (YOLOX, RFDETR, RTMDet) and a pose estimator, plus utilities for drawing skeletons. Models are auto‑downloaded from OpenMMLab or a HuggingFace mirror. Install via pip or from source, and run a Gradio WebUI or use the Python API for single‑image, video, or webcam inference.