#4040th of 75 in ML frameworks and libraries
kornia
Differentiable computer vision and geometry operators for PyTorch
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 11.4k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- May 2026
- Language
- Python
11,097 on 31 Mar 202611,440 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Kornia is a PyTorch library of over 500 differentiable image processing, augmentation and geometry ops, including filters, homographies, camera models, epipolar geometry and Lie groups. It also ships pretrained models such as LoFTR, LightGlue, SAM and YuNet, plus an ONNXSequential helper for chaining ONNX models. The README publishes per-module float16/bfloat16 test status and known failures.
Who it is for: PyTorch developers doing vision, augmentation or geometry work
Strengths
- Batched, differentiable ops that run on GPU tensors inside PyTorch training pipelines
- Covers filters, augmentation, feature matching, cameras, epipolar geometry and Lie groups in one library
- Publishes CPU and CUDA test pass rates and known half-precision failures per module
- Installs with pip install kornia; Apache-2.0 licensed
Weaknesses
- Half precision is incomplete: about 95% pass in float16 and bfloat16 on CPU
- Epipolar solvers raise NotImplementedError in half precision on CPU
- No CI job covers CUDA or MPS half precision; CUDA results are a local run
- Roadmap favors conformance and benchmarks over new API growth
What it needs
- GPU optional
- Needs PyTorch 2.5.1+, kornia_rs (used in quick start), onnx (optional extra)
- Models: PyTorch, ONNX, LoFTR, LightGlue, SAM
Also in ML frameworks and libraries
See all 75| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 92 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 91 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 90 out of 100 |