Skip to content

RELEASE_NOTES.md

Latest commit

 

History

History
104 lines (76 loc) · 3.03 KB

RELEASE_NOTES.md

File metadata and controls

104 lines (76 loc) · 3.03 KB

SQPnP v2.0.0 - Python Bindings Release

🎉 What's New

This major release adds comprehensive Python bindings to the SQPnP library, making it accessible to a wider audience of researchers and developers.

✨ New Features

Python Bindings

  • Easy-to-use Python API: Simple interface with NumPy arrays
  • Multiple camera models: Support for pinhole, fisheye, and distortion models
  • Integrated RANSAC: Robust estimation with outlier rejection
  • High performance: Direct C++ bindings with minimal overhead
  • Real-time demo: Live head pose estimation with webcam

Documentation & Examples

  • Comprehensive README: Updated with both C++ and Python usage
  • Detailed API reference: Complete documentation for all features
  • Installation scripts: Automated setup for different platforms
  • Real-time demo: Head pose estimation with MediaPipe integration
  • Performance benchmarks: Comprehensive testing and optimization

Developer Experience

  • Automated installation: Cross-platform install script
  • Comprehensive testing: Unit tests and integration tests
  • Error handling: Robust validation and error reporting
  • Performance optimization: Optimized for real-time applications

🔧 Installation

C++ (Existing)

mkdir build && cd build
cmake ..
make -j4

Python (New)

cd python_bindings
pip install -e .

📖 Quick Start

Python Usage

import numpy as np
import sqpnp_python

# Create solver and solve PnP
solver = sqpnp_python.SQPnPSolver()
result = solver.solve_pinhole(points_3d, points_2d, camera_params)

if result.success:
    print(f"Rotation: {result.rotation}")
    print(f"Translation: {result.translation}")

Real-time Demo

python adapter.py

🎮 Demo Features

  • Real-time face detection using MediaPipe
  • 3D head pose estimation using SQPnP
  • Visual axes overlay on the face
  • Performance metrics (FPS, CPU usage)
  • Support for different landmark subsets

📊 Performance

  • C++: Optimized for embedded devices and real-time applications
  • Python: Minimal overhead with direct C++ bindings
  • RANSAC: Integrated outlier rejection for robust estimation
  • Multi-threading: Support for parallel processing

🔄 Migration Guide

No breaking changes for existing C++ users. Python users can start with the new bindings immediately.

📚 Documentation

🙏 Acknowledgments

  • Original SQPnP implementation by G. Terzakis and M. Lourakis
  • RANSAC implementation from RansacLib
  • Python bindings using pybind11
  • Real-time demo using MediaPipe Face Mesh

📄 License

Same as the original SQPnP project.

Note: This release maintains full backward compatibility with existing C++ code while adding powerful new Python capabilities.