easy-yolo 0.1.2

Easy to use library for YOLO inference in rust requiring no additional setup, weights included.
Documentation

A small library for running YOLO models from rust using ort. Supports YOLO-NAS exported from super gradients and ultralytics YOLOv8/11/12.

Run the example:

cargo run --release --features example --example run_yolo /path/to/an/image.jpeg

Yolo V12 nano weights are included in the libary, allowing you to run YOLO without any setup or extra downloads. The weights are contained in a zip file inside the binary (to get around crates.io 10MB limit), and unzipped at runtime using the zip library. This funtionality is behind the 'weights' feature flag.

TODOs:

  • Refine and improve support for different model output formats
  • Refactor to not be fixed to 640x640 resolution everywhere
  • Improve ort feature handling, for enabling more runtimes (curerntly only have coreML)
  • Train/find a quantised model with weights small enough to fit in the 10MB cargo crate size limit, so the library can be used without any setup DONE! I found Yolo V12 nano weights and zipped them.