void Start()
{
runtimeModel = ModelLoader.Load(model);
worker = WorkerFactory.CreateWorker(runtimeModel);
#if UNITY_EDITOR
if (testImages.pixelValueArray.Length == 0)
{
testImages.LoadBytesFromPath();
}
if (testNetwork)
{
int fails = 0;
Tensor input = new Tensor(1, 0, 28 * 28, 1);
for (int i = 0; i < testImages.imageCount; i++)
{
for (int j = 0; j < testImages.pixelValueArrayOffset; j++)
{
input[0, 0, j, 0] =
testImages.pixelValueArray[i * testImages.pixelValueArrayOffset + j];
}
Tensor output = worker.Execute(input).PeekOutput("Y");
int value = GetResult(output);
bool correctGuess = value == testImages.labelValueArray[i];
fails += correctGuess ? 0 : 1;
}
Debug.Log(fails + " fails of " + testImages.imageCount + " images");
Debug.Log((1 - ((float)fails / testImages.imageCount)) * 100 + "% accuracy");
}
#endif
TestRandomImage(true);
}
// Start is called before the first frame update
void Start()
{
if (trainedNetwork == null || (trainedNetwork != null && !trainedNetwork.networkSet))
{
//load data
trainingImages.LoadBytesFromPath();
//Create training data array
trainingData = new TrainingData[trainingImages.imageCount];
uint pixelCount = trainingImages.pixelWidth * trainingImages.pixelHeight;
for (int i = 0; i < trainingImages.imageCount; i++)
{
trainingData[i].input = new float[11]; //input nodes
trainingData[i].targetResult = new float[pixelCount]; //target nodes
trainingData[i].input[10] = RandomGaussian(); //Noise
//Set data
trainingData[i].input[trainingImages.labelValueArray[i]] = 1f;
for (int j = 0; j < pixelCount; j++)
{
trainingData[i].targetResult[j] = trainingImages.pixelValueArray[i * trainingImages.pixelValueArrayOffset + j];
}
}
// (inputNodeCount, hiddenNodeCount, outputNodeCount)
neuralNetwork = new NeuralNetwork_Matrix(
trainingData[0].input.Length,
(trainingData[0].input.Length + trainingData[0].targetResult.Length) / 2,
trainingData[0].targetResult.Length);
neuralNetwork.TrainNeuralNetwork(trainingData);
if (trainedNetwork != null)
{
trainedNetwork.SetNetworkVariables(neuralNetwork);
}
}
else
{
neuralNetwork = new NeuralNetwork_Matrix(trainedNetwork);
}
Texture2D tempTex = new Texture2D(28, 28);
image.sprite = Sprite.Create(tempTex, new Rect(0, 0, 28, 28), Vector2.zero);
CreateImange();
}
// Start is called before the first frame update
void Start()
{
if (trainedNetwork == null || (trainedNetwork != null && !trainedNetwork.networkSet))
{
//load data
trainingImages.LoadBytesFromPath();
//Create training data array
trainingData = new TrainingData[trainingImages.imageCount];
uint pixelCount = trainingImages.pixelWidth * trainingImages.pixelHeight;
for (int i = 0; i < trainingImages.imageCount; i++)
{
trainingData[i].input = new float[pixelCount]; //input nodes
trainingData[i].targetResult = new float[11]; //target nodes
//Set data
trainingData[i].targetResult[trainingImages.labelValueArray[i]] = 1f;
trainingData[i].targetResult[10] = 1f;
for (int j = 0; j < pixelCount; j++)
{
trainingData[i].input[j] = trainingImages.pixelValueArray[i * trainingImages.pixelValueArrayOffset + j];
}
}
// (inputNodeCount, hiddenNodeCount, outputNodeCount)
discriminator = new NeuralNetwork_Matrix(
trainingData[0].input.Length,
(trainingData[0].input.Length + trainingData[0].targetResult.Length) / 2,
trainingData[0].targetResult.Length);
generator = new NeuralNetwork_Matrix(11, (11 + 28 * 28) / 2, 28 * 28);
//if (trainedNetwork != null)
//{
// trainedNetwork.SetNetworkVariables(discriminator);
//}
}
//else
//{
// discriminator = new NeuralNetwork_Matrix(trainedNetwork);
//}
testImages.LoadBytesFromPath();
TrainNetwork(1);
}
// Start is called before the first frame update
void Start()
{
guessText.text = "A";
if (trainedNetwork == null || (trainedNetwork != null && !trainedNetwork.networkSet))
{
trainingImages.LoadBytesFromPath();
guessText.text = "B";
//Create training data array
trainingData = new TrainingData[trainingImages.imageCount + NOISE_TRAINING_AMOUNT];
uint pixelCount = trainingImages.pixelWidth * trainingImages.pixelHeight;
guessText.text = "C";
for (int i = 0; i < trainingImages.imageCount; i++)
{
trainingData[i].input = new float[pixelCount]; //input nodes
trainingData[i].targetResult = new float[11]; //target nodes
//Set data
trainingData[i].targetResult[trainingImages.labelValueArray[i]] = 1f;
for (int j = 0; j < pixelCount; j++)
{
trainingData[i].input[j] = trainingImages.pixelValueArray[i * trainingImages.pixelValueArrayOffset + j];
}
}
for (int i = 0; i < NOISE_TRAINING_AMOUNT; i++)
{
trainingData[trainingImages.imageCount + i].input = new float[pixelCount]; //input nodes
trainingData[trainingImages.imageCount + i].targetResult = new float[11]; //target nodes
trainingData[trainingImages.imageCount + i].targetResult[10] = 1f;
for (int j = 0; j < pixelCount; j++)
{
trainingData[trainingImages.imageCount + i].input[j] = Random.Range(0f, 1f);
}
}
guessText.text = "D";
// (inputNodeCount, hiddenNodeCount, outputNodeCount)
neuralNetwork = new NeuralNetwork_Matrix(
trainingData[0].input.Length,
(trainingData[0].input.Length + trainingData[0].targetResult.Length) / 2,
trainingData[0].targetResult.Length);
guessText.text = "E";
neuralNetwork.TrainNeuralNetwork(trainingData, 70000, true);
guessText.text = "F";
if (trainedNetwork != null)
{
trainedNetwork.SetNetworkVariables(neuralNetwork);
}
guessText.text = "G";
}
else
{
neuralNetwork = new NeuralNetwork_Matrix(trainedNetwork);
}
guessText.text = "H";
#if UNITY_EDITOR
if (testImages.pixelValueArray.Length == 0)
{
testImages.LoadBytesFromPath();
}
#endif
if (testNetwork)
{
TestNeuralNetwork();
}
guessText.text = "I";
TestRandomImage();
}