/*Method changed*/
public async Task PredImages <T>(T directory, CancellationToken token)
{
IEnumerable <string> files;
if (directory is string dirPath)
{
files = await Task.Run <IEnumerable <string> >(() =>
{ return(Directory.EnumerateFiles(dirPath)); });
}
else if (directory is IEnumerable <string> dirFiles)
{
files = dirFiles;
}
else
{
throw new ArgumentException("Only types string and IEnumerable<string> are supported");
}
ParallelOptions options = new ParallelOptions();
options.CancellationToken = token;
Parallel.ForEach(files, options, (file) =>
{
MNISTModelResult result = new MNISTModelResult(file);
result = this.PredImage(file);
ResultIsReady?.Invoke(this, new ResultEventArgs(result));
});
}
public MNISTModelResult PredImage <T>(T inputImage)
{
Image <Rgb24> image = null;
MNISTModelResult result = new MNISTModelResult();
IImageFormat format = null;
string imagePath = null;
if (inputImage is string)
{
imagePath = inputImage as string;
image = Image.Load <Rgb24>(imagePath, out format);
}
else if (inputImage is byte[] byteArr)
{
image = Image.Load <Rgb24>(byteArr, out format);
}
//crop
Stream imageStream = new MemoryStream();
image.Mutate(x =>
{
x.Resize(new ResizeOptions
{
Size = new Size(imageWidth, imageHeight),
Mode = ResizeMode.Crop
});
});
image.Save(imageStream, format);
//make a tensor
Tensor <float> input = new DenseTensor <float>(new[] { 1, 1, image.Width, image.Height });
for (int y = 0; y < image.Height; y++)
{
Span <Rgb24> pixelSpan = image.GetPixelRowSpan(y);
for (int x = 0; x < image.Width; x++)
{
input[0, 0, y, x] = (0.299f * pixelSpan[x].R / 255f
+ 0.587f * pixelSpan[x].G / 255f
+ 0.114f * pixelSpan[x].B / 255f);
}
}
//make an input
var inputs = new List <NamedOnnxValue>
{
NamedOnnxValue.CreateFromTensor(InputName, input)
};
//run model
var session = new InferenceSession(ModelPath);
var results = session.Run(inputs);
//postprocess output
IEnumerable <float> output = results.First().AsEnumerable <float>();
IEnumerable <float> softmax = output.Select(x => (float)Math.Exp(x) /
output.Sum(x => (float)Math.Exp(x)));
//select max
Dictionary <int, float> dictOutput = new Dictionary <int, float>();
int key = 0;
foreach (var value in softmax)
{
dictOutput.Add(key, value);
key++;
}
var maxConfElem = dictOutput.FirstOrDefault(elem => elem.Value == dictOutput.Values.Max());
result.Path = imagePath;
result.Class = maxConfElem.Key;
result.Confidence = maxConfElem.Value;
return(result);
}
public ResultEventArgs(MNISTModelResult result)
{
Result = new MNISTModelResult(result.Path, result.Class, result.Confidence);
}
public MNISTModelResult(MNISTModelResult result)
{
Path = result.Path;
Class = result.Class;
Confidence = result.Confidence;
}