readme.md

.NET libraries for Lobe

A .NET library to run inference on exported Lobe models.

How to get started

Export the model from Lobe app

• Export your model as ONNX format

Use the model in your own .NET application

Install the follwing packages

• lobe.Onnx to import the Onnx based implementation of the image classifier.
• lobe.ImageSharp to get image manipulation utilities
• Microsoft.ML.OnnxRuntime to get he native onnx runtimes

This code creates a simple command line app that loads a model and classifies an image file

using System;
using System.IO;
using SixLabors.ImageSharp;
using SixLabors.ImageSharp.PixelFormats;
using lobe.ImageSharp;

namespace lobe.TestApp
{
    class Program
    {
        static void Main(string[] args)
        {
            var signatureFilePath = args[0];
            var imageToClassify = args[1];

            ImageClassifier.Register("onnx", () => new OnnxImageClassifier());
            using var classifier = ImageClassifier.CreateFromSignatureFile(
                new FileInfo(signatureFilePath));

            var results = classifier.Classify(Image
                .Load(imageToClassify).CloneAs<Rgb24>());
            Console.WriteLine(results.Classification.Label);
        }
    }
}

The code

ImageClassifier.Register("onnx", () => new OnnxImageClassifier());

Registers the OnnxImageClassifier against the format onnx

Then a classifier can be built from a signature file

using var classifier = ImageClassifier.CreateFromSignatureFile(new FileInfo(signatureFilePath));

Use the lobe app directly

For rapid iterations you can test your model by taking advantage of the http endpoint that the lobe app exposes. First, open the app and then the model you would like to use. Next, go to the export menu and select the api option, get the url from there (it should look like http://localhost:38100/predict/bdff75cc-ee54-46cf-a290-f9095ef78516").

In your .NET App make sure you have installed the following nuget packages

• lobe.Http
• lobe.ImageSharp

Then import the namespaces

using System.IO;
using SixLabors.ImageSharp;

using lobe.Http;

And finally create the client, connect to the endpoint and now use it classify the image

var client = new LobeClient();
client.UseUri(new Uri("http://localhost:38100/predict/bdff75cc-ee54-46cf-a290-f9095ef78516"));

var result = client.Classify(picture.CloneAs<Rgb24>());