Ultralytics YOLO Rust Examples

Runnable examples for the ultralytics-inference library. Each example is a single file you run with cargo run --example. They show how to load and run Ultralytics YOLO26, Ultralytics YOLO11, and Ultralytics YOLOv8 models from Rust.

📂 Examples

Example	Feature	Description	Run
basic	none	Load a model, run inference, print detections	cargo run --example basic
config	none	Set confidence, IoU, image size, and device	cargo run --example config
tasks	none	Summary for every task, plus raw arrays for segment, semantic, depth	cargo run --example tasks
annotate	annotate	Draw boxes and labels, save the annotated image	cargo run --example annotate --features annotate

✅ How to Run

Run any example from the repository root. The model and a sample image are downloaded on first use, so a network connection is needed once.

# Print detections for the sample image
cargo run --example basic

# Use your own image
cargo run --example basic -- path/to/image.jpg

# Set thresholds, image size, and device
cargo run --example config

# Annotate and save to runs/predict/annotated_*.jpg
cargo run --example annotate --features annotate

The tasks example defaults to detection and works with the other tasks when you pass their model. The matching sample image is downloaded automatically. The segment, semantic, and depth branches also print the raw output array, which ndarray truncates with ...:

cargo run --example tasks                       # detect (default)
cargo run --example tasks -- yolo26n-seg.onnx   # segment
cargo run --example tasks -- yolo26n-pose.onnx  # pose
cargo run --example tasks -- yolo26n-obb.onnx   # obb
cargo run --example tasks -- yolo26n-cls.onnx   # classify
cargo run --example tasks -- yolo26n-sem.onnx   # semantic (Ultralytics YOLO26)
cargo run --example tasks -- yolo26n-depth.onnx # depth (Ultralytics YOLO26)

The basic example prints one block per image:

Found 5 detections
  bus 0.93 [5.9 228.8 807.3 748.9]
  person 0.92 [46.7 398.6 237.0 901.8]
  person 0.90 [223.0 405.3 345.2 863.1]
  person 0.86 [668.6 391.2 810.0 879.7]
  person 0.51 [0.0 553.4 65.4 874.1]

The examples load yolo26n.onnx, downloading it on first use. To provide the model yourself, export it with the Ultralytics Python package. The command below writes yolo26n.onnx into the current directory, which the example then loads instead of downloading. The positional argument is still the input image:

pip install ultralytics
yolo export model=yolo26n.pt format=onnx

cargo run --example basic -- image.jpg

To run a different model, change the model path in the example.

🤝 Contributing

Contributions are welcome. See the Ultralytics Contributing Guide for details. If you find an issue with an example or want to add one, open an issue or pull request on the Ultralytics inference repository.