Introduction

Welcome to Iris! This library provides high-performance computer vision algorithms, image manipulation tools, and deep learning model hosting fully in safe Rust.

NOTE

This project is still in active development. APIs and features may change before the first stable release.

Why Iris?

Traditional computer vision libraries are often large C++ codebases. Using them from Rust requires binding wrappers, which introduce:

• Complicated build toolchains, especially on Windows or bare-metal environments.
• Linking conflicts with system-level dynamic libraries.
• Unsafe memory pointers that bypass Rust's borrow checker.

Iris takes a different approach — every algorithm is implemented natively in Rust:

• Memory Safe: All code runs within Rust's ownership system. No null pointer dereferences, no buffer overflows, no data races.
• Type Safe: Strong compile-time type guarantees prevent shape mismatches, invalid tensor operations, and API misuse.
• Zero-Dependency Compilation: Clean Cargo builds with no external C/C++ system libraries required.
• Modern Hardware Acceleration: Run the same operations on CPU, WGPU (Vulkan, DX12, Metal), or CUDA by switching the backend type parameter.
• Rayon Multi-Threading: CPU-intensive loops scale automatically to all available cores.

Backend Generic Design

Every Iris function is generic over the Burn Backend trait. This means you can write your code once and run it on any supported backend:

use iris::prelude::*;
use burn::backend::wgpu::Wgpu;

type Backend = Wgpu;
let device = Default::default();
let img = Image::<Backend>::open("photo.jpg", &device)?;

Switch to NdArray for CPU-only environments:

use burn::backend::ndarray::NdArray;
type Backend = NdArray;