# kornia/kornia

Kornia: A Differentiable Computer Vision Library for Spatial AI

- HackerByte page: <https://hackerbyte.net/en/projects/kornia-a-differentiable-computer-vision-library-for-spatial-ai-145693916>
- GitHub: <https://github.com/kornia/kornia>
- Programming language: Python

🐍 Geometric Computer Vision Library for Spatial AI

## AI-generated summary

Kornia is a Python library built on PyTorch that offers a comprehensive set of differentiable image processing and geometric computer vision algorithms. It integrates seamlessly with AI workflows, providing tools for batch transformations, auto\-differentiation, and GPU acceleration. The library aims to be a reference implementation for differentiable computer vision and geometry within the PyTorch ecosystem.

## Key points

- Offers over 1,200 public operators, many unique compared to torchvision, covering areas like camera models, epipolar geometry, feature detection/matching, and image losses.
- Supports per\-sample parameters in single batched calls for transformations and enhancements, unlike torchvision's single\-parameter approach.
- Includes batched minimal solvers and robust estimation methods like RANSAC, with options for device\-based computation and compilation.
- Features differentiable geometry operations for 3D transformations, camera models, and depth processing, maintaining gradients with respect to pose and camera parameters.
- Provides a range of AI models for tasks such as face detection, feature matching, segmentation, and classification, alongside advanced augmentation pipelines.