X, y = training_generator.__next__()
print('X,y created')
# zero center images
X = np.array(X)
#X = X/255
print('X as arrray created')
# Fitting the model. Here you can see how the call to model fit works. Note the validation data comes from
# preloaded numpy arrays.
print('one hot encoding ...')
y = to_categorical(y)
#if iteration != 0:
# model.load_weights(check_point_weights)
history = model.fit(
[X[0], X[1]],
y,
batch_size=16,
epochs=30,
validation_data=([validationLeft, validationRight], validationLabels),
callbacks=[checkpointer])
histories.append(history)
show(histories, False)
#tuned = load_model(check_point_model)
#result = tuned.evaluate([validationLeft, validationRight], validationLabels)
#print(result)
#Define directories
baseDir = r"D:\Arnaud\data_croutinet\ottawa\data"
trainDir = os.path.join(baseDir, "train/train.csv")
validationDir = os.path.join(baseDir, "validation/validation.csv")
base_network_save = os.path.join(baseDir, "scoreNetworkRetrain2.h5")
ranking_network_save = os.path.join(baseDir, "rankingNetworkRetrain.h5")
base_network_save2 = os.path.join(baseDir, "scoreNetworkRetrain3.h5")
#load training and validation set with labels as scalars between 0 and 1
trainLeft, trainRight, trainLabels = loadAsScalars(trainDir)
validationLeft, validationRight, validationLabels = loadAsScalars(
validationDir)
#Here is the architecture of ScoreCroutinet that we create below
base_network = load_model(base_network_save)
model = create_meta_network(INPUT_DIM, base_network)
#We fit the model to the training set
history = model.fit([trainLeft, trainRight],
trainLabels,
batch_size=16,
epochs=30,
validation_data=([validationLeft,
validationRight], validationLabels))
#We show the result and save the network
show([history], False)
base_network.save(base_network_save2)