YOLOv8

Use YOLOv8 in real-time for object detection, instance segmentation, pose estimation and image classification, via ONNX Runtime

Install

The YoloV8 project is available in two nuget packages: YoloV8 and YoloV8.Gpu, if you use with CPU add the YoloV8 package reference to your project (contains reference to Microsoft.ML.OnnxRuntime package)

dotnet add package YoloV8

If you use with GPU you need to add the YoloV8.Gpu package reference (contains reference to Microsoft.ML.OnnxRuntime.Gpu package)

dotnet add package YoloV8.Gpu

Use

Export the model from PyTorch to ONNX format:

Run the following python code to export the model to ONNX format:

from ultralytics import YOLO

# Load a model
model = YOLO('path/to/best')

# export the model to ONNX format
model.export(format='onnx')

Use in exported model with C#:

using Compunet.YoloV8;
using SixLabors.ImageSharp;

using var predictor = YoloV8Predictor.Create("path/to/model");

var result = predictor.Detect("path/to/image");
// or
var result = await predictor.DetectAsync("path/to/image");

Console.WriteLine(result);

Plotting

You can to plot the input image for preview the model prediction results, this code demonstrates how to perform a prediction with the model and then plot the prediction results on the input image and save to file:

using Compunet.YoloV8;
using Compunet.YoloV8.Plotting;
using SixLabors.ImageSharp;

var imagePath = "path/to/image";

using var predictor = YoloV8Predictor.Create("path/to/model");

var result = await predictor.PoseAsync(imagePath);

using var image = Image.Load(imagePath);
using var ploted = await result.PlotImageAsync(image);

ploted.Save("./pose_demo.jpg")

Demo Images:

Detection:

Pose:

Segmentation:

License

MIT License