private void saveMLPState(MultiLayeredPerceptronClassifier mlp, float accuracy)
{
MLPState mstate = new MLPState()
{
inputNeuronsArr = mlp.inputNeuronsArr,
hiddenNeuronsArr = mlp.hiddenNeuronsArr,
outputNeuronsArr = mlp.outputNeuronsArr,
weightsArr = mlp.weightsArr,
tempWeightsArr = mlp.tempWeightsArr,
prevWeightsArr = mlp.prevWeightsArr,
inputs = mlp.inputs,
outputs = mlp.outputs,
hiddenLayers = mlp.hiddenLayers,
hiddenNeurons = mlp.hiddenNeurons,
meanSqErr = mlp.meanSqErr,
epochs = mlp.epochs,
error = mlp.error
};
// write the data (overwrites)
using (var stream = new StreamWriter(MLP_STATE_FILE, append: false)) /// + "_" + accuracy
{
stream.Write(JsonConvert.SerializeObject(mstate));
}
}
private void loadMLPState()
{
using (var stream = new StreamReader(MLP_STATE_FILE))
{
MLPState mstate = JsonConvert.DeserializeObject <MLPState>(stream.ReadToEnd());
// Load state from file.
MultiLayeredPerceptronClassifier.CreateFromFile(mstate);
}
}
/**
* Multilayer perceptron function.
*
* @param trainDatasetLimit: Apply limit on training dataset for each character 1 - 275 or -1 for MAX.
* @throws Exception
*/
private float runMLP(int trainDatasetLimit)
{
List <OCRCharacter> charactersTrain = new List <OCRCharacter>();
List <OCRCharacter> charactersTest = new List <OCRCharacter>();
// Load the training file points
loadDataFromFile(charactersTrain, DATASET_FILE_1, DATASET_FILE_CONTENT_1);
// Load the training file points from dataset 2 as well.
loadDataFromFile(charactersTest, DATASET_FILE_2, DATASET_FILE_CONTENT_2);
int limit = trainDatasetLimit <= 0 ? 280 : trainDatasetLimit;
List <OCRCharacter> charactersTrainFiltered =
ResultFilter.Instance.limitResults(charactersTrain, limit);
Stopwatch stopwatch = new Stopwatch();
stopwatch.Start();
bool saveState = true;
// Load ready config from file if present.
if (System.IO.File.Exists(MLP_STATE_FILE))
{
loadMLPState();
saveState = false;
}
// Create a static singleton instance
mlp = MultiLayeredPerceptronClassifier.Instance;
if (!mlp.isTrained)
{
mlp.createNetwork();
Console.WriteLine("Found number of images: " + charactersTrainFiltered.Count + "\n");
mlp.trainNetwork(0.04f, 0.01f, 0.4f, charactersTrainFiltered, 500);
}
int correct = 0;
foreach (OCRCharacter entry in charactersTest)
{
correct += mlp.recallNetwork(entry) ? 1 : 0;
}
stopwatch.Stop();
long estimatedTime = stopwatch.ElapsedMilliseconds; //System.nanoTime() - startTime;
// Show accuracy results.
Console.WriteLine(correct + " / " + charactersTest.Count);
float accuracy = (float)(correct * 100) / charactersTest.Count;
Console.WriteLine(accuracy + " % accuracy.");
// Time elapsed
Console.WriteLine("Execution time is " + ((float)estimatedTime / 1000000000).ToString("#0.000") + " seconds");
// Save to file, to avoid long load time.
if (saveState)
{
saveMLPState(mlp, accuracy);
}
/// saveMLPState(mlp, accuracy);
return(accuracy);
}
