def main_loop():
print("fitting the whole model ")
model = minc_model1()
model.fit(X_train,
Y_train,
batch_size=batch_size,
nb_epoch=nb_epoch,
verbose=1,
validation_data=(X_test, Y_test))
model.summary()
score = model.evaluate(X_test, Y_test, verbose=0)
print('Test score:', score[0])
print('Test accuracy:', score[1])
# Save the model to another location.
output_name = 'model_saved/minc_cnn_sndstat.weights'
out_dir = get_absolute_dir_project(output_name)
print('saving model to location -> {} '.format(out_dir))
model.save_weights(out_dir)
model2 = model_without_dense()
model2.load_weights(out_dir, by_name=True)
model2.summary()
model2.fit(X_train,
Y_train,
batch_size=batch_size,
nb_epoch=nb_epoch,
verbose=1,
validation_data=(X_test, Y_test))
score = model2.evaluate(X_test, Y_test, verbose=0)
print('Test score:', score[0])
print('Test accuracy:', score[1])
def evaluate_model():
model = mnist_model1()
output_name = 'model_saved/mnist_cnn_sndstat.weights'
out_dir = get_absolute_dir_project(output_name)
model.load_weights(out_dir)
score = model.evaluate(X_test, Y_test, verbose=0)
print('Test score:', score[0])
print('Test accuracy:', score[1])
from keras.utils import np_utils
from keras.utils.data_utils import get_absolute_dir_project
from kyu.utils.logger import Logger
import sys
batch_size = 32
nb_classes = 20
nb_epoch = 1000
data_augmentation = True
# input image dimensions
img_rows, img_cols = 32, 32
# the CIFAR10 images are RGB
img_channels = 3
BASELINE_PATH = get_absolute_dir_project(
'model_saved/cifar10_baseline.weights')
SND_PATH = get_absolute_dir_project('model_saved/cifar10_cnn_sndstat.weights')
LOG_PATH = get_absolute_dir_project('model_saved/log')
# the data, shuffled and split between train and test sets
# label_mode = 'fine'
label_mode = 'coarse'
(X_train, y_train), (X_test,
y_test) = cifar100.load_data(label_mode=label_mode)
if label_mode is 'fine':
nb_classes = 100
elif label_mode is 'coarse':
nb_classes = 20
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
from keras.utils.data_utils import get_absolute_dir_project
from kyu.datasets.minc import Minc2500
from kyu.utils.example_engine import ExampleEngine
LOG_PATH = get_absolute_dir_project('model_saved/log/minc2500')
# global constants
NB_CLASS = 23 # number of classes
LEARNING_RATE = 0.01
MOMENTUM = 0.9
BATCH_SIZE = 128
ALPHA = 0.0001
BETA = 0.75
GAMMA = 0.1
DROPOUT = 0.5
WEIGHT_DECAY = 0.0005
NB_EPOCH = 20
LRN2D_norm = True # whether to use batch normalization
# Theano - 'th' (channels, width, height)
# Tensorflow - 'tf' (width, height, channels)
DIM_ORDERING = 'th'
TARGET_SIZE = (224, 224)
### FOR model 1
if K.backend() == 'tensorflow':
INPUT_SHAPE = TARGET_SIZE + (3, )
K.set_image_dim_ordering('tf')
else:
INPUT_SHAPE = (3, ) + TARGET_SIZE
K.set_image_dim_ordering('th')
def test_loader():
# Save the model to another location.
output_name = 'model_saved/mnist_cnn_sndstat.weights'
out_dir = get_absolute_dir_project(output_name)
print('saving model to location -> {} '.format(out_dir))
