nb_dense_block=nb_dense_block,
growth_rate=growth_rate,
nb_filter=nb_filter,
dropout_rate=dropout_rate,
weights=None)
print('Model created')
model.summary()
model.load_weights('densenet-filter20.hdf5', by_name=True)
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-3),
metrics=['accuracy'])
print('Model compiled')
x = resource.load_train_x(t=30)
y = resource.load_train_y()
(trainX, testX) = (x[:int(50000 * 0.9)], x[int(50000 * 0.9):])
(trainY, testY) = (y[:int(50000 * 0.9)], y[int(50000 * 0.9):])
trainX = trainX.astype('float32')
testX = testX.astype('float32')
mean_image = np.mean(trainX, axis=0)
trainX -= mean_image
testX -= mean_image
trainX /= 128.0
testX /= 128.0
trainY = np_utils.to_categorical(trainY, nb_classes)
testY = np_utils.to_categorical(testY, nb_classes)
generator = ImageDataGenerator(rotation_range=15,
width_shift_range=5./64,
nb_dense_block=nb_dense_block,
growth_rate=growth_rate,
nb_filter=nb_filter,
dropout_rate=dropout_rate,
weights=None)
print('Model created')
model.summary()
model.load_weights('densenet-filter200.hdf5', by_name=True)
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-3),
metrics=['accuracy'])
print('Model compiled')
x = resource.load_train_x(t=255)
y = resource.load_train_y()
(trainX, testX) = (x[:int(50000 * 0.9)], x[int(50000 * 0.9):])
(trainY, testY) = (y[:int(50000 * 0.9)], y[int(50000 * 0.9):])
trainX = trainX.astype('float32')
testX = testX.astype('float32')
mean_image = np.mean(trainX, axis=0)
trainX -= mean_image
testX -= mean_image
trainX /= 128.0
testX /= 128.0
trainY = np_utils.to_categorical(trainY, nb_classes)
testY = np_utils.to_categorical(testY, nb_classes)
generator = ImageDataGenerator(rotation_range=15,
width_shift_range=5. / 64,
img_rows, img_cols = 64, 64
img_channels = 1
img_dim = (img_channels, img_rows, img_cols)
model = resnet.ResnetBuilder.build_resnet_50(img_dim, nb_classes)
print('Model created')
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-3),
metrics=['accuracy'])
print('Model compiled')
x = resource.load_train_x()
y = resource.load_train_y()
(trainX, testX) = (x[:int(50000 * 0.9)], x[int(50000 * 0.9):])
(trainY, testY) = (y[:int(50000 * 0.9)], y[int(50000 * 0.9):])
trainX = trainX.astype('float32')
testX = testX.astype('float32')
mean_image = np.mean(trainX, axis=0)
trainX -= mean_image
testX -= mean_image
trainX /= 128.0
testX /= 128.0
trainY = np_utils.to_categorical(trainY, nb_classes)
testY = np_utils.to_categorical(testY, nb_classes)
generator = ImageDataGenerator(rotation_range=15,
width_shift_range=5. / 64,
nb_dense_block=nb_dense_block,
growth_rate=growth_rate,
nb_filter=nb_filter,
dropout_rate=dropout_rate,
weights=None)
print('Model created')
model.summary()
model.load_weights('densenet-filter70.hdf5', by_name=True)
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-3),
metrics=['accuracy'])
print('Model compiled')
x = resource.load_train_x(t=100)
y = resource.load_train_y()
(trainX, testX) = (x[:int(50000 * 0.9)], x[int(50000 * 0.9):])
(trainY, testY) = (y[:int(50000 * 0.9)], y[int(50000 * 0.9):])
trainX = trainX.astype('float32')
testX = testX.astype('float32')
mean_image = np.mean(trainX, axis=0)
trainX -= mean_image
testX -= mean_image
trainX /= 128.0
testX /= 128.0
trainY = np_utils.to_categorical(trainY, nb_classes)
testY = np_utils.to_categorical(testY, nb_classes)
generator = ImageDataGenerator(rotation_range=15,
width_shift_range=5. / 64,
