/// <summary>
/// SetupPsi() is called at application startup. It is responsible for
/// building and starting the Psi pipeline
/// </summary>
public void SetupPsi()
{
Console.WriteLine("================================================================================");
Console.WriteLine(" Kiosk Awareness sample");
Console.WriteLine("================================================================================");
Console.WriteLine();
this.pipeline = Pipeline.Create();
// Next register an event handler to catch pipeline errors
this.pipeline.PipelineCompletionEvent += this.PipelineCompletionEvent;
/*
* // bool usingKqml = false;
* string facilitatorIP = null;
* int facilitatorPort = -1;
* int localPort = -1;
*
*
* if (arguments.Length > 0)
* {
* if (arguments.Length < 3)
* {
* Console.WriteLine("Usage for running with a facilitator: \nKioskMain facilitatorIP facilitatorPort localPort");
* return;
* }
*
* // usingKqml = true;
*
* facilitatorIP = arguments[0];
* facilitatorPort = int.Parse(arguments[1]);
* localPort = int.Parse(arguments[2]);
* }
*/
// bool showLiveVisualization = false;
string inputLogPath = null;
// string outputLogPath = null;
DateTime startTime = DateTime.Now;
IProducer <AudioBuffer> audioInput = SetupAudioInput(this.pipeline, inputLogPath, ref startTime);
// Create our webcam
MediaCapture webcam = new MediaCapture(this.pipeline, 320, 240, 10);
Debug.WriteLine("Open webcam");
FaceCasClassifier f = new FaceCasClassifier();
Debug.WriteLine("Load classifier");
Debug.WriteLine(f);
var mouthOpenAsBool = webcam.Out.ToGrayViaOpenCV(f, FrameCount).Select(
(img, e) =>
{
// Debug.WriteLine(FrameCount % 10);
bool mouthOpen = false;
if ((Math.Abs(DisNose) / (4 * Math.Abs(DisLipMiddle))) < 3)
{
mouthOpen = true;
}
else
{
mouthOpen = false;
}
Console.WriteLine(Math.Abs(DisLipMiddle) + " " + Math.Abs(DisLipRight) + " " + Math.Abs(DisLipLeft) + " " + (Math.Abs(DisNose) / (4 * Math.Abs(DisLipMiddle))) + " " + mouthOpen);
this.DispImage.UpdateImage(img);
return(mouthOpen);
});
var mouthAndSpeech = audioInput.Pair(mouthOpenAsBool).Where(t => true).Select(t =>
{
return(t.Item1);
}
);
SystemSpeechRecognizer recognizer = SetupSpeechRecognizer(this.pipeline);
mouthAndSpeech.PipeTo(recognizer);
var finalResults = recognizer.Out.Where(result => result.IsFinal);
finalResults.Do(result =>
{
var ssrResult = result as SpeechRecognitionResult;
Console.WriteLine($"{ssrResult.Text} (confidence: {ssrResult.Confidence})");
});
var text = finalResults.Select(result =>
{
var ssrResult = result as SpeechRecognitionResult;
return(ssrResult.Text);
});
// Finally start the pipeline running
try
{
this.pipeline.RunAsync();
}
catch (AggregateException exp)
{
MessageBox.Show("Error! " + exp.InnerException.Message);
}
}
/// <summary>
/// Here we define an Psi extension. This extension will take a stream of images (source)
/// and create a new stream of converted images.
/// </summary>
/// <param name="source">Our source producer (source stream of image samples)</param>
/// <param name="f">A wapper face classifier object (null means use the default)</param>
/// <param name="framecount">A integer to control the frame number</param>
/// <param name="deliveryPolicy">Our delivery policy (null means use the default)</param>
/// <returns>The new stream of converted images.</returns>
public static IProducer <Shared <Image> > ToGrayViaOpenCV(this IProducer <Shared <Image> > source, FaceCasClassifier f = null, int framecount = 0, DeliveryPolicy deliveryPolicy = null)
{
// Process informs the pipeline that we want to call our lambda ("(srcImage, env, e) =>{...}") with each image
// from the stream.
return(source.Process <Shared <Image>, Shared <Image> >(
(srcImage, env, e) =>
{
// Our lambda here is called with each image sample from our stream and calls OpenCV to convert
// the image into a grayscale image. We then post the resulting gray scale image to our event queu
// so that the Psi pipeline will send it to the next component.
// Have Psi allocate a new image. We will convert the current image ('srcImage') into this new image.
using (var destImage = ImagePool.GetOrCreate(srcImage.Resource.Width, srcImage.Resource.Height, PixelFormat.Gray_8bpp))
{
// Call into our OpenCV wrapper to convert the source image ('srcImage') into the newly created image ('destImage')
// Note: since srcImage & destImage are Shared<> we need to access the Microsoft.Psi.Imaging.Image data via the Resource member
OpenCVMethods.ToGray(srcImage.ToImageBuffer(), destImage.ToImageBuffer(), f, ref MainWindow.DisNose, ref MainWindow.DisLipMiddle, ref MainWindow.DisLipRight, ref MainWindow.DisLipLeft, ref MainWindow.HasFace);
// Debug.WriteLine(MainWindow.MouthOpen);
e.Post(destImage, env.OriginatingTime);
}
}, deliveryPolicy));
}
public static void StartListeningAndLooking(string[] args, bool live_visual_flag, bool store_visual_flag, string inputStorePath, string outputStorePath, bool usingKqml, String[] compargs)
{
using (Pipeline pipeline = Pipeline.Create())
{
string facilitatorIP = null;
int facilitatorPort = -1;
int localPort = -1;
/*
* if (args.Length > 0)
* {
* if (args.Length < 3)
* {
* Console.WriteLine("Usage for running with a facilitator: \nKioskMain facilitatorIP facilitatorPort localPort");
* return;
* }
* usingKqml = true;
*
* facilitatorIP = args[0];
* facilitatorPort = int.Parse(args[1]);
* localPort = int.Parse(args[2]);
* }
*/
string outputLogPath = null;
if (outputStorePath != null && outputStorePath != "" && Directory.Exists(outputStorePath))
{
outputLogPath = outputStorePath;
}
Console.WriteLine(outputLogPath == null);
string inputLogPath = null;
if (inputStorePath != null && inputStorePath != "" && Directory.Exists(inputStorePath))
{
inputLogPath = inputStorePath;
}
Console.WriteLine(inputLogPath == null);
bool showLiveVisualization = live_visual_flag;
// Needed only for live visualization
DateTime startTime = DateTime.Now;
// Use either live audio from the microphone or audio from a previously saved log
IProducer <AudioBuffer> audioInput = SetupAudioInput(pipeline, inputLogPath, ref startTime);
// Create our webcam
MediaCapture webcam = new MediaCapture(pipeline, 320, 240, 10);
FaceCasClassifier f = new FaceCasClassifier();
Console.WriteLine("Load classifier");
Console.WriteLine(f);
// Bind the webcam's output to our display image.
// The "Do" operator is executed on each sample from the stream (webcam.Out), which are the images coming from the webcam
var processedVideo = inputLogPath != null?SetupVideoInput(pipeline, inputLogPath, ref startTime) : webcam.Out.ToGrayViaOpenCV(f).EncodeJpeg(90, DeliveryPolicy.LatestMessage);
var mouthOpenAsInt = processedVideo.Select(
(img, e) =>
{
// Debug.WriteLine(FrameCount % 10);
// Console.WriteLine(Math.Abs(DisLipMiddle) + " " + Math.Abs(DisLipRight) + " " + Math.Abs(DisLipLeft) + " " + (Math.Abs(DisNose) / (4 * Math.Abs(DisLipMiddle))) + " " + mouthOpen);
//return MouthOpen;
return(MouthOpen);
});
/*
* var hasFaceAsBool = webcam.Out.ToGrayViaOpenCV(f).Select(
* (img, e) =>
* {
* bool hasFacebool = false;
* if (HasFace == 1)
* {
* hasFacebool = true;
* }
* else
* {
* hasFacebool = false;
* }
* return hasFacebool;
* });
*/
var mouthAndSpeech = audioInput.Pair(mouthOpenAsInt).Where(t => t.Item2 > -1).Select(t => {
return(t.Item1);
}
);
SystemSpeechRecognizer recognizer = SetupSpeechRecognizer(pipeline);
// Subscribe the recognizer to the input audio
mouthAndSpeech.PipeTo(recognizer);
//audioInput.PipeTo(recognizer);
// Partial and final speech recognition results are posted on the same stream. Here
// we use Psi's Where() operator to filter out only the final recognition results.
var finalResults = inputLogPath != null?SetupSpeechInput(pipeline, inputLogPath, ref startTime) : recognizer.Out.Where(result => result.IsFinal);
// Print the recognized text of the final recognition result to the console.
finalResults.Do(result =>
{
var ssrResult = result as SpeechRecognitionResult;
Console.WriteLine($"{ssrResult.Text} (confidence: {ssrResult.Confidence})");
});
var finalResultsHighCf = finalResults.Where(t => (t as SpeechRecognitionResult).Confidence > 0.6).Select(t =>
{
Console.WriteLine("Good Confidence!");
return(t);
});
// Get just the text from the Speech Recognizer. We may want to add another filter to only get text if confidence > 0.8
var text = finalResultsHighCf.Pair(mouthOpenAsInt).Select(result =>
{
var ssrResult = result.Item1 as SpeechRecognitionResult;
int userid = result.Item2;
Console.WriteLine("user" + userid + "+" + ssrResult.Text);
return("user" + userid + "+" + ssrResult.Text);
});
// Setup KQML connection to Companion
NU.Kqml.SocketStringConsumer kqml = null;
if (usingKqml)
{
facilitatorIP = compargs[0];
facilitatorPort = Convert.ToInt32(compargs[1]);
localPort = Convert.ToInt32(compargs[2]);
Console.WriteLine("Your Companion IP address is: " + facilitatorIP);
Console.WriteLine("Your Companion port is: " + facilitatorPort);
Console.WriteLine("Your local port is: " + localPort);
kqml = new NU.Kqml.SocketStringConsumer(pipeline, facilitatorIP, facilitatorPort, localPort);
text.PipeTo(kqml.In);
}
// Create a data store to log the data to if necessary. A data store is necessary
// only if output logging or live visualization are enabled.
Console.WriteLine(outputLogPath == null);
var dataStore = CreateDataStore(pipeline, outputLogPath, showLiveVisualization);
Console.WriteLine(dataStore == null);
Console.WriteLine("dataStore is empty");
// For disk logging or live visualization only
if (dataStore != null)
{
// Log the microphone audio and recognition results
processedVideo.Write($"{Program.AppName}.WebCamProcessedVideo", dataStore);
audioInput.Write($"{Program.AppName}.MicrophoneAudio", dataStore);
finalResults.Write($"{Program.AppName}.FinalRecognitionResults", dataStore);
Console.WriteLine("Stored the data here! ");
}
// Ignore this block if live visualization is not enabled
if (showLiveVisualization)
{
// Create the visualization client
var visualizationClient = new VisualizationClient();
// Clear all data if the visualizer is already open
visualizationClient.ClearAll();
// Create the visualization client to visualize live data
visualizationClient.SetLiveMode(startTime);
// Plot the video stream in a new panel
visualizationClient.AddXYPanel();
//processedVideo.Show(visualizationClient);
// Plot the microphone audio stream in a new panel
visualizationClient.AddTimelinePanel();
//audioInput.Show(visualizationClient);
// Plot the recognition results in a new panel
visualizationClient.AddTimelinePanel();
//finalResults.Show(visualizationClient);
}
if (store_visual_flag)
{
// Create the visualization client
var visualizationClient = new VisualizationClient();
// Clear all data if the visualizer is already open
visualizationClient.ClearAll();
// Create the visualization client to visualize live data
visualizationClient.SetLiveMode(startTime);
// Plot the video stream in a new panel
visualizationClient.AddXYPanel();
processedVideo.Show(visualizationClient);
// Plot the microphone audio stream in a new panel
visualizationClient.AddTimelinePanel();
audioInput.Show(visualizationClient);
// Plot the recognition results in a new panel
visualizationClient.AddTimelinePanel();
finalResults.Show(visualizationClient);
}
// Register an event handler to catch pipeline errors
pipeline.PipelineCompletionEvent += PipelineCompletionEvent;
// Run the pipeline
pipeline.RunAsync();
Console.WriteLine("Press any key to exit...");
Console.ReadKey(true);
// if (kqml != null) kqml.Stop();
}
}
