def load_dataset_tensors():
"""
Retonar os tensors para serem treinados com o Keras
"""
print("Carregando os dados...")
train_files, targets_train = get_dataset("train")
valid_files, targets_valid = get_dataset("val")
# pre-process the data for Keras
print("Convertendo para tensor...")
train_tensors = paths_to_tensor(train_files)
valid_tensors = paths_to_tensor(valid_files)
return train_tensors, targets_train, valid_tensors, targets_valid
def load_train_tensors():
"""
Retonar os tensors para serem treinados com o Keras (Só treinamento)
"""
print("Carregando os dados...")
train_files, targets_train = get_dataset("train", True)
# pre-process the data for Keras
print("Convertendo para tensor...")
train_tensors = paths_to_tensor(train_files)
return train_tensors, targets_train
def load_dataset_tensors():
"""
Carrega as imagens e converte para tenors para treinar a deep
"""
files1 = util.get_all_files_names("output/Dia/")
files2 = util.get_all_files_names("output/Noite/")
files3 = util.get_all_files_names("output/Trasnsicao/")
files_names = files1 + files2 + files3
labels = [0] * len(files1) + [1] * len(files2) + [2] * len(files3)
files_train, files_val, targets_train, targets_valid = train_test_split(
files_names, labels, test_size=.2)
train_tensors = util.paths_to_tensor(files_train)
train_tensors = preprocess_input(train_tensors)
valid_tensors = util.paths_to_tensor(files_val)
valid_tensors = preprocess_input(valid_tensors)
targets_train = np_utils.to_categorical(util.np.array(targets_train), 3)
targets_valid = np_utils.to_categorical(util.np.array(targets_valid), 3)
return train_tensors, targets_train, valid_tensors, targets_valid
def load_test_tensors(val=False):
"""
Retonar os tensors de test
"""
print("Carregando os dados...")
if val:
test_files, targets_test = get_dataset("val", True)
else:
test_files, targets_test = get_dataset("test", True)
# pre-process the data for Keras
print("Convertendo para tensor...")
test_tensors = paths_to_tensor(test_files)
return test_tensors, targets_test
def load_dataset_tensors(nexet=False):
"""
Carrega as imagens e converte para tenors para treinar a deep
"""
if nexet:
files_names = pdn.get_all_files_nexet()[0:10000]
else:
files1 = util.get_all_files_names("output/Dia/")
files2 = util.get_all_files_names("output/Noite/")
files3 = util.get_all_files_names("output/Trasnsicao/")
files_names = files1 + files2 + files3
x = util.paths_to_tensor(files_names)
x = pi(x)
return x
def get_CNN_features(files_names=None):
"""
Extrai features com CNNs pre-treinadas e os retorna
"""
model = ResNet50(include_top=False, weights='imagenet')
# model = InceptionV3(include_top=False, weights='imagenet')
# model = Xception(include_top=False, weights='imagenet')
if files_names is None:
files1 = util.get_all_files_names("output/Dia/")
files2 = util.get_all_files_names("output/Noite/")
files3 = util.get_all_files_names("output/Trasnsicao/")
files_names = files1 + files2 + files3
tensors = util.paths_to_tensor(files_names)
tensors = preprocess_input(tensors)
pred = model.predict(tensors).reshape(len(files_names), 2048)
return pred
for i_file in [0]:
files = [train_files, valid_files][i_file]
target = [train_targets, valid_targets][i_file]
print("Incializando o pre data Augmentation...")
datagen_train = ImageDataGenerator(
rotation_range=40,
width_shift_range=0.2,
height_shift_range=0.2,
# rescale=1./255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
fill_mode='nearest')
tensors = paths_to_tensor(files)
datagen_train.fit(tensors)
j = 0
for X_batch, y_batch in datagen_train.flow(
tensors, target, batch_size=n_examples[i_file]):
print("Gerando batch %d" % j)
count = int(np.ceil(len(X_batch) / block_size))
print(len(X_batch), count)
# for i in range(count):
# print("{0} lotte: {1}/{2}".format(files_names[i_file][0], i, count))
# begin = i*block_size
# end = begin + block_size
# if end > len(X_batch):
# end = len(X_batch)
from keras.callbacks import ModelCheckpoint
from keras.preprocessing.image import load_img, img_to_array, ImageDataGenerator
from PIL import ImageFile
from joblib import Parallel, delayed
from util import paths_to_tensor
import numpy as np
import multiprocessing
if __name__ == "__main__":
print("Carregando os dados...")
valid_files, valid_targets = get_dataset("val")
# pre-process the data for Keras
print("Convertendo para tensor...")
valid_tensors = paths_to_tensor(valid_files).astype('float32') / 255.0
model = Sequential()
model.add(
ResNet50(weights='imagenet',
include_top=False,
input_shape=(224, 224, 3)))
model.add(Flatten(name='flatten'))
model.add(Dense(83, activation='softmax'))
model.compile(optimizer='rmsprop',
loss='binary_crossentropy',
metrics=['accuracy'])
### Load the model weights with the best validation loss.
model.load_weights('saved_models/weights.best.ResNet50.hdf5')
train_files, train_targets = get_dataset("train")
valid_files, valid_targets = get_dataset("val")
files_names = ["dataset_train3.csv", "dataset_val3.csv"]
block_size = 500
# 0 - train, 1 - validação
for i_file in [0, 1]:
files = [train_files, valid_files][i_file]
count = int(np.ceil(len(files) / block_size))
for i in range(count):
print("{0} lotte: {1}/{2}".format(files_names[i_file], i, count))
begin = i * block_size
end = begin + block_size
if end > len(files):
end = len(files)
# Processa um lote de imagens paralelamente
tensors = paths_to_tensor(files[begin:end])
tensors = preprocess_input(tensors)
pred = model.predict(tensors).reshape(end - begin, 2048)
# Adiciona essas imagens ao csv
df(pred).to_csv(path + files_names[i_file],
mode='a',
header=(i == 10),
index=False)