Overview

Retinify is an advanced AI-powered stereo vision library designed for robotics.
It enables real-time, high-precision 3D perception by leveraging GPU and NPU acceleration.

Note

The source code is publicly available:
👉 View on GitHub

The retinify pipeline

The main functionality of retinify is accessible through retinify::Pipeline.

GPU-Based Computing:
All processing required for stereo depth estimation — including image remapping (undistortion and rectification), stereo matching, and 3D reprojection — is executed entirely on the GPU, enabling efficient and real-time performance.

AI-Powered Stereo Matching:
Stereo correspondence is computed using deep learning–based models, delivering significantly higher accuracy and robustness compared to traditional stereo matching algorithms.

Flexible Integration:
The pipeline is camera-agnostic and supports any pinhole stereo camera setup. Input images are handled via raw pointers and strides, allowing seamless integration without dependencies on specific image processing libraries.

Getting started

• Installation
  
• Demos
  
• Tutorials
  
• Calibration
• Python Docs
  
• C++ Docs
  
• ROS2 Docs
  

Contact

For all inquiries, including support, collaboration, please contact:
contact@retinify.ai