edgefirst-hal

EdgeFirst Hardware Abstraction Layer - A unified Rust library for edge AI inference pipelines.

This is the umbrella crate that re-exports the core EdgeFirst HAL components:

• edgefirst-tensor - Zero-copy tensor memory management (DMA, SHM, system memory)
• edgefirst-image - High-performance image processing and format conversion
• edgefirst-decoder - ML model output decoding (YOLO, ModelPack)

Features

• Zero-copy memory management with DMA-BUF and POSIX shared memory support
• Hardware-accelerated image processing via G2D (NXP i.MX) and OpenGL
• Efficient ML post-processing for object detection and segmentation models
• Cross-platform - Linux (with hardware acceleration), macOS, and other Unix systems

Quick Start

use edgefirst_hal::{tensor, image, decoder};

// Create a tensor with automatic memory selection
let tensor = tensor::Tensor::<f32>::new(&[1, 3, 640, 640], None, None)?;

// Load and process an image
let img = image::TensorImage::load(&image_bytes, Some(image::RGBA), None)?;
let mut processor = image::ImageProcessor::new()?;
processor.convert(&img, &mut dst, image::Rotation::None, image::Flip::None, image::Crop::default())?;

// Decode YOLO model outputs
let decoder = decoder::DecoderBuilder::new()
    .with_score_threshold(0.25)
    .with_iou_threshold(0.7)
    .build()?;

Platform Support

License

Licensed under the Apache License, Version 2.0. See LICENSE for details.

Links

• EdgeFirst AI
• GitHub Repository
• Documentation