def trainCNNModel(X_LL_train, X_LH_train, X_HL_train, X_HH_train, y_train,
X_LL_test, X_LH_test, X_HL_test, X_HH_test, y_test,
num_epochs):
batch_size = 32 # in each iteration, we consider 32 training examples at once
num_train, height, width, depth = X_LL_train.shape
num_classes = len(np.unique(y_train))
Y_train = np_utils.to_categorical(y_train,
num_classes) # One-hot encode the labels
Y_test = np_utils.to_categorical(y_test,
num_classes) # One-hot encode the labels
checkPointFolder = 'checkPoint'
checkpoint_name = checkPointFolder + '/Weights-{epoch:03d}--{val_loss:.5f}.hdf5'
checkpoint = ModelCheckpoint(checkpoint_name,
monitor='val_loss',
verbose=1,
save_best_only=True,
mode='auto')
callbacks_list = [checkpoint]
if not os.path.exists(checkPointFolder):
os.makedirs(checkPointFolder)
model = createModel(height, width, depth, num_classes)
model.compile(
loss='binary_crossentropy',
#loss='categorical_crossentropy', # using the cross-entropy loss function
optimizer='adam', # using the Adam optimiser
metrics=['accuracy']) # reporting the accuracy
model.fit(
[X_LL_train, X_LH_train, X_HL_train, X_HH_train],
Y_train, # Train the model using the training set...
batch_size=batch_size,
epochs=num_epochs,
verbose=1,
validation_split=0.1,
callbacks=callbacks_list
) # ...holding out 10% of the data for validation
score, acc = model.evaluate(
[X_LL_test, X_LH_test, X_HL_test, X_HH_test], Y_test,
verbose=1) # Evaluate the trained model on the test set!
model.save('moirePattern3CNN_.h5')
return model
def main(args):
weights_file = (args.weightsFile)
positiveImagePath = (args.positiveTestImages)
negativeImagePath = (args.negativeTestImages)
os.system("python3 createTrainingData.py {} {} {}".format(
positiveImagePath, negativeImagePath, 1))
X_LL, X_LH, X_HL, X_HH, X_index, Y, imageCount = readWaveletData(
positiveImagePath, negativeImagePath, positiveTestImagePath,
positiveTestImagePath)
X_LL = np.array(X_LL)
X_LL = X_LL.reshape((imageCount, height, width, depth))
X_LH = np.array(X_LH)
X_LH = X_LH.reshape((imageCount, height, width, depth))
X_HL = np.array(X_HL)
X_HL = X_HL.reshape((imageCount, height, width, depth))
X_HH = np.array(X_HH)
X_HH = X_HH.reshape((imageCount, height, width, depth))
CNN_model = createModel(height, width, depth, num_classes)
CNN_model.load_weights(weights_file)
evaluate(CNN_model, X_LL, X_LH, X_HL, X_HH, Y)