wellton/Emosic
1
0
Fork 0
Código Issues Pull Requests Ações Pacotes Projetos Versões Wiki Atividade

update net

Ver Código-Fonte
Esse commit está contido em:
Charlie Hewitt
2017-12-14 22:28:05 +00:00
pai 9c4dddff30
commit 0f711a7afa
1 arquivos alterados com 55 adições e 49 exclusões
Baixar Arquivo de Patch Baixar Arquivo de Diff Expandir todos os arquivos Colapsar todos os arquivos
net.py
+55 -49
Ver Arquivo
@@ -1,6 +1,6 @@
from keras.preprocessing import image
from keras.models import Sequential
from keras.layers import Conv2D, Conv3D, MaxPooling2D, MaxPooling3D, Activation, Dropout, Flatten, Dense
from keras.layers import Conv2D, Conv3D, MaxPooling2D, MaxPooling3D, Activation, Dropout, Flatten, Dense, BatchNormalization, AveragePooling2D
from keras.utils.np_utils import to_categorical
from keras.optimizers import SGD, Adam
from keras.callbacks import LearningRateScheduler, ModelCheckpoint
@@ -16,8 +16,8 @@ REGRESS = 1
# OTPIONS #
CLASSIFY_OR_REGRESS = CLASSIFY
BATCH_SIZE = 32
EPOCHS = 8
BATCH_SIZE = 256
EPOCHS = 16
# LOADERS #
def load_images(paths, labels, batch_size=32):
@@ -29,16 +29,16 @@ def load_images(paths, labels, batch_size=32):
if batch_n*batch_size + i > len(paths) - 1:
batch_n = 0
path = paths[batch_n*batch_size + i]
img = image.load_img(path, target_size=(256, 256))
img = image.load_img(path, target_size=(96, 96))
x = image.img_to_array(img)/255
# x = image.random_rotation(x, 10)
# x = image.random_shift(x, 0.1, 0.1)
# if np.random.random() < 0.5:
# x = image.flip_axis(x, 1)
x = image.random_rotation(x, 20)
x = image.random_shift(x, 0.1, 0.1)
if np.random.random() < 0.5:
x = image.flip_axis(x, 1)
y = labels[batch_n*batch_size + i]
batch_d.append(x)
batch_l.append(y)
batch_d = np.array(batch_d).reshape((batch_size, 256, 256, 3))
batch_d = np.array(batch_d).reshape((batch_size, 96, 96, 3))
if CLASSIFY_OR_REGRESS == CLASSIFY:
batch_l = np.array(batch_l).reshape((batch_size, 8))
else:
@@ -92,39 +92,53 @@ def process_data(paths, labels):
count += 1
print('Processed:', count, end='\r')
if CLASSIFY_OR_REGRESS == CLASSIFY:
weights = class_weight.compute_class_weight('balanced', np.unique(labels_out), labels_out)
weights = dict(enumerate(weights))
labels_out = to_categorical(labels_out, num_classes=8)
else:
weights = None
print('Processed:', count)
return paths_out, labels_out
return paths_out, labels_out, weights
# MODEL #
model = Sequential()
# TODO sort out model
model.add(Conv2D(32, (16, 16), input_shape=(256, 256, 3), activation='relu'))
model.add(Conv2D(32, (16, 16), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.1))
model.add(Conv2D(64, (8, 8), activation='relu'))
model.add(Conv2D(64, (8, 8), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.1))
model.add(Conv2D(128, (4, 4), activation='relu'))
model.add(Conv2D(128, (4, 4), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))
model.add(Conv2D(256, (3, 3), activation='relu'))
model.add(Conv2D(256, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))
model.add(Conv2D(512, (3, 3), activation='relu'))
model.add(Conv2D(512, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
# CONV BLOCK 1
model.add(Conv2D(32, (3, 3), input_shape=(96, 96, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(Conv2D(32, (3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
# CONV BLOCK 2
model.add(Conv2D(64,(3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(Conv2D(64, (3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
# CONV BLOCK 3
model.add(Conv2D(128,(3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(Conv2D(128, (3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
# FLATTEN
model.add(Flatten())
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.4))
model.add(Dense(4096, activation='relu'))
model.add(Dropout(0.4))
# sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
# DENSE 1
model.add(Dense(128))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.5))
# DENSE 2
model.add(Dense(128))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.5))
# OUTPUT
if CLASSIFY_OR_REGRESS == CLASSIFY:
model.add(Dense(8, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
@@ -135,28 +149,20 @@ else:
print('** LOADING DATA **')
t_paths = np.load('training_paths.npy')
t_labels = np.load('training_labels.npy')
t_paths, t_labels = process_data(t_paths, t_labels)
t_paths, t_labels, t_weights = process_data(t_paths, t_labels)
v_paths = np.load('validation_paths.npy')
v_labels = np.load('validation_labels.npy')
v_paths, v_labels = process_data(v_paths, v_labels)
v_paths, v_labels, v_weights = process_data(v_paths, v_labels)
print('** FITTING MODEL **')
# def lr_schedule(epoch):
# if epoch >= 1:
# lr = 0.01 * (1 - epoch/EPOCHS)
# else:
# lr = 0.1
# return lr
model.fit_generator(
load_images(t_paths, t_labels, BATCH_SIZE),
steps_per_epoch=len(t_labels)//BATCH_SIZE,
class_weight='auto',
class_weight=t_weights,
epochs=EPOCHS,
validation_data=load_images(v_paths, v_labels, BATCH_SIZE),
validation_steps=len(v_labels),
callbacks=[#LearningRateScheduler(lr_schedule),
ModelCheckpoint('AFF_NET_'+str(CLASSIFY_OR_REGRESS)+'WIP.h5', save_best_only=True)])
validation_steps=len(v_labels)//BATCH_SIZE,
callbacks=[ModelCheckpoint('AFF_NET_'+str(CLASSIFY_OR_REGRESS)+'_WIP.h5', save_best_only=True)])
print('** EXPORTING MODEL **')
for k in model.layers: