title | description | parent | grand_parent | has_children | redirect_from | nav_order |
---|---|---|---|---|---|---|
Object detection with Faster RCNN in C# | The sample walks through how to run a pretrained Faster R-CNN object detection ONNX model using the ONNX Runtime C# API. | Inference with C# | Tutorials | false | /docs/tutorials/fasterrcnn_csharp | 2 |
{: .no_toc }
The sample walks through how to run a pretrained Faster R-CNN object detection ONNX model using the ONNX Runtime C# API.
The source code for this sample is available here.
{: .no_toc }
- TOC placeholder {:toc}
To run this sample, you'll need the following things:
- Install .NET Core 3.1 or higher for you OS (Mac, Windows or Linux).
- Download the Faster R-CNN ONNX model to your local system.
- Download this demo image to test the model. You can also use any image you like.
Now we have everything set up, we can start adding code to run the model on the image. We'll do this in the main method of the program for simplicity.
Firstly, let's read the path to the model, path to the image we want to test, and path to the output image:
stringmodelFilePath=args[0];stringimageFilePath=args[1];stringoutImageFilePath=args[2];
Next, we will read the image in using the cross-platform image library ImageSharp:
usingImage<Rgb24>image=Image.Load<Rgb24>(imageFilePath,outIImageFormatformat);
Note, we're specifically reading the Rgb24
type so we can efficiently preprocess the image in a later step.
Next, we will resize the image to the appropriate size that the model is expecting; it is recommended to resize the image such that both height and width are within the range of [800, 1333].
floatratio=800f/Math.Min(image.Width,image.Height);usingStreamimageStream=newMemoryStream();image.Mutate(x =>x.Resize((int)(ratio*image.Width),(int)(ratio*image.Height)));image.Save(imageStream,format);
Next, we will preprocess the image according to the requirements of the model:
varpaddedHeight=(int)(Math.Ceiling(image.Height/32f)*32f);varpaddedWidth=(int)(Math.Ceiling(image.Width/32f)*32f);varmean=new[]{102.9801f,115.9465f,122.7717f};// Preprocessing image// We use DenseTensor for multi-dimensional accessDenseTensor<float>input=new(new[]{3,paddedHeight,paddedWidth});image.ProcessPixelRows(accessor =>{for(inty=paddedHeight-accessor.Height;y<accessor.Height;y++){Span<Rgb24>pixelSpan=accessor.GetRowSpan(y);for(intx=paddedWidth-accessor.Width;x<accessor.Width;x++){input[0,y,x]=pixelSpan[x].B-mean[0];input[1,y,x]=pixelSpan[x].G-mean[1];input[2,y,x]=pixelSpan[x].R-mean[2];}}});
Here, we're creating a Tensor of the required size (channels, paddedHeight, paddedWidth)
, accessing the pixel values, preprocessing them and finally assigning them to the tensor at the appropriate indices.
// Pin DenseTensor memory and use it directly in the OrtValue tensor // It will be unpinned on ortValue disposal
usingvarinputOrtValue=OrtValue.CreateTensorValueFromMemory(OrtMemoryInfo.DefaultInstance,input.Buffer,newlong[]{3,paddedHeight,paddedWidth});
Next, we will create the inputs to the model:
varinputs=newDictionary<string,OrtValue>{{"image",inputOrtValue}};
To check the input node names for an ONNX model, you can use Netron to visualize the model and see input/output names. In this case, this model has image
as the input node name.
Next, we will create an inference session and run the input through it:
usingvarsession=newInferenceSession(modelFilePath);usingvarrunOptions=newRunOptions();usingIDisposableReadOnlyCollection<OrtValue>results=session.Run(runOptions,inputs,session.OutputNames);
Next, we will need to postprocess the output to get boxes and associated label and confidence scores for each box:
varboxesSpan=results[0].GetTensorDataAsSpan<float>();varlabelsSpan=results[1].GetTensorDataAsSpan<long>();varconfidencesSpan=results[2].GetTensorDataAsSpan<float>();constfloatminConfidence=0.7f;varpredictions=newList<Prediction>();for(inti=0;i<boxesSpan.Length-4;i+=4){varindex=i/4;if(confidencesSpan[index]>=minConfidence){predictions.Add(newPrediction{Box=newBox(boxesSpan[i],boxesSpan[i+1],boxesSpan[i+2],boxesSpan[i+3]),Label=LabelMap.Labels[labelsSpan[index]],Confidence=confidencesSpan[index]});}}
Note, we're only taking boxes that have a confidence above 0.7 to remove false positives.
Next, we'll draw the boxes and associated labels and confidence scores on the image to see how the model went:
usingvaroutputImage=File.OpenWrite(outImageFilePath);Fontfont=SystemFonts.CreateFont("Arial",16);foreach(varpinpredictions){image.Mutate(x =>{x.DrawLines(Color.Red,2f,newPointF[]{newPointF(p.Box.Xmin,p.Box.Ymin),newPointF(p.Box.Xmax,p.Box.Ymin),newPointF(p.Box.Xmax,p.Box.Ymin),newPointF(p.Box.Xmax,p.Box.Ymax),newPointF(p.Box.Xmax,p.Box.Ymax),newPointF(p.Box.Xmin,p.Box.Ymax),newPointF(p.Box.Xmin,p.Box.Ymax),newPointF(p.Box.Xmin,p.Box.Ymin)});x.DrawText($"{p.Label}, {p.Confidence:0.00}",font,Color.White,newPointF(p.Box.Xmin,p.Box.Ymin));});}image.Save(outputImage,format);
For each box prediction, we're using ImageSharp to draw red lines to create the boxes, and drawing the label and confidence text.
Now the program is created, we can run it will the following command:
dotnet run [path-to-model] [path-to-image] [path-to-output-image]
e.g. running:
dotnet run ~/Downloads/FasterRCNN-10.onnx ~/Downloads/demo.jpg ~/Downloads/out.jpg
detects the following objects in the image: