🦀
Pure idiomatic Rust
Written natively in Rust from the ground up. Zero unsafe wrappers, zero external C/C++ dependencies.
IrisNative Rust Computer Vision
A fast computer vision library in pure Rust.
Skip to content
NOTE
This project is still in active development. APIs and features may change before the first stable release.
Iris Documentation
Welcome to the official documentation for Iris — a pure-Rust computer vision library. Explore the guides below to get started:
Getting Started
Install Iris in your project, load an image, apply a Gaussian blur, detect edges with Canny, and save the result — all in a few lines of Rust.
Installation
Cargo features, backend options (WGPU, CUDA, LibTorch, Ndarray), and build configuration. Customize Iris for your hardware and use case.
Image Representation
How images are represented as Burn tensors with shape [Channels, Height, Width]. Create, load, save, and query image properties.
Image Filters & Blur
Box blur, Gaussian blur, median filter, bilateral filter, and separable 2D filtering — all parallelized with Rayon or accelerated on GPU.
Edge Detection
Canny edge detection, Sobel, Scharr, and Laplacian gradient operators for structural analysis and feature extraction.
DNN & ONNX Inference
Load and run ONNX, Safetensors, and PyTorch bin models. Preprocess inputs with blob_from_image and filter results with NMS.
API Reference
Full reference for all types, modules, and functions — Image, Point, Rect, Scalar, filters, edges, morphology, DNN, and GUI.