def process_input(self, Y):
'''
:param string y: Input STRING
'''
start = time()
# Save the raw input string
raw_input = Y
Y = self.vectorizer.transform([Y])
y_tf_idf = self.tf_idf_transformer.transform(Y)
distances, ids = self.nbrs.kneighbors(y_tf_idf.todense())
# Rate questions by their similarity scores using w2v
q_similarity_scores = {}
input_tokens = preprocess_input(raw_input)
for id in ids[0]:
question = self.dataset[id][0]
sum_similarities = 0
question = preprocess_input(question)
for word in input_tokens:
for q_word in question:
try:
# Find the maximum similarity of each input word
# to each word in the question
sims = [self.w2v_model.wv.similarity(word, q_word)]
sum_similarities += max(sims)
except KeyError:
print(f"Word {q_word} not in dataset")
# Associate every question with it's similarity
# to the input
q_similarity_scores[id] = sum_similarities
# Sort question IDs by their similarity to the input
sorted_by_sim = {id: sim for id, sim in sorted(q_similarity_scores.items(), key = lambda x: x[1], reverse=True) }
# Return the top 10 results
retval = []
i = 0
for res in sorted_by_sim:
retval.append((res, self.dataset[res]))
i += 1
if i == 10:
break
return retval
def telemetry(sid, data):
if data:
# The current steering angle of the car
steering_angle = data["steering_angle"]
# The current throttle of the car
throttle = data["throttle"]
# The current speed of the car
speed = data["speed"]
# The current image from the center camera of the car
imgString = data["image"]
image = Image.open(BytesIO(base64.b64decode(imgString)))
image_array = np.asarray(image)
steering_angle = float(
model.predict(preprocessing.preprocess_input(
image_array[None, :, :, :]),
batch_size=1))
throttle = controller.update(float(speed))
print(steering_angle, throttle)
send_control(steering_angle, throttle)
# save frame
if args.image_folder != '':
timestamp = datetime.utcnow().strftime('%Y_%m_%d_%H_%M_%S_%f')[:-3]
image_filename = os.path.join(args.image_folder, timestamp)
image.save('{}.jpg'.format(image_filename))
else:
# NOTE: DON'T EDIT THIS.
sio.emit('manual', data={}, skip_sid=True)
def set_dataset(self, dataset, processed_dataset, corpus):
'''
:param dict dataset: {key, [question, answer]}
'''
self.dataset = dataset
start = time()
list_of_q = self._extract_questions(dataset)
self.vectorizer = CountVectorizer(lowercase=True, analyzer='word')
X = self.vectorizer.fit_transform(list_of_q)
self.tf_idf_transformer = TfidfTransformer(use_idf=True).fit(X)
x_tf_idf = self.tf_idf_transformer.transform(X)
self.nbrs = NearestNeighbors(n_neighbors=10, algorithm='ball_tree', metric='manhattan').fit(x_tf_idf)
print(start - time())
self.corpus = corpus
# Extract token list per question
# for the w2v model
token_list = []
for question in list_of_q:
token_list.append(preprocess_input(question))
# Train w2v model
self.w2v_model = Word2Vec(token_list, size=100, window=3, min_count=0, workers=4, iter=10)
def main(_):
if not os.path.exists("./outputs/steering_model"):
os.makedirs("./outputs/steering_model")
batch_size = 512
X_image = []
X_flip = []
y_steering = []
with open('driving_data.csv', 'rt') as csvfile:
reader = csv.reader(csvfile, skipinitialspace=True)
for row in reader:
X_image.append(row[0])
X_flip.append(float(row[1]))
y_steering.append(float(row[2]))
X_image = np.array(X_image)
X_flip = np.array(X_flip)
y_steering = np.array(y_steering)
X_train_image, X_test_image, X_train_flip, X_test_flip, y_train_steering, y_test_steering = train_test_split(X_image, X_flip, y_steering, test_size=0.2)
if len(X_train_image) == 0:
print('No data found')
return
sample_image = preprocessing.preprocess_input(np.array([mpimg.imread(X_train_image[0])]))[0]
print('Image Shape: ', sample_image.shape)
print('Training Data Count:', len(X_train_image))
print('Validation Data Count:', len(X_test_image))
X_train_image, X_train_flip, y_train_steering = shuffle(X_train_image, X_train_flip, y_train_steering)
comma_ai_model = get_comma_ai_model(sample_image.shape)
nvidia_model = get_nvidia_model(sample_image.shape)
merged = Merge([comma_ai_model, nvidia_model])
model = Sequential()
model.add(merged)
model.add(Dense(1, activation='elu'))
model.compile(optimizer="adam", loss="mse")
model.fit_generator(image_generator(X_train_image, X_train_flip, y_train_steering, batch_size), samples_per_epoch=1024*25, nb_epoch=5, validation_data=image_generator(X_test_image, X_test_flip, y_test_steering, batch_size), nb_val_samples=1024 * 5)
print("Saving model weights and configuration file.")
model.save_weights("./outputs/steering_model/steering_angle.keras", True)
with open('./outputs/steering_model/steering_angle.json', 'w') as outfile:
json.dump(model.to_json(), outfile)
def ExpressionDetection (image):
detection_model_path = "%s/%s/haarcascade_frontalface_default.xml"%(PATH,"haar_cascade")
emotion_model_path = '%s/vision/models/fer2013_mini_XCEPTION.102-0.66.hdf5'%(PATH)
emotion_labels = get_labels('fer2013')
# loading models
face_detection = load_detection_model(detection_model_path)
emotion_classifier = load_model(emotion_model_path, compile=False)
# getting input model shapes for inference
emotion_target_size = emotion_classifier.input_shape[1:3]
emotion_offsets = (20, 40)
#read image
# print 'dddddddd',gray_image
#gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# rgb_image = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gray_image=cv2.imread(image,0)
faces = detect_faces(face_detection,gray_image)
# print 'qqqqqqqq',faces
for face_coordinates in faces:
x1, x2, y1, y2 = apply_offsets(face_coordinates, emotion_offsets)
gray_face = gray_image[y1:y2, x1:x2]
try:
gray_face = cv2.resize(gray_face, (emotion_target_size))
except:
continue
gray_face = preprocess_input(gray_face, True)
# print 'eeeeee',gray_face
gray_face = np.expand_dims(gray_face, 0)
gray_face = np.expand_dims(gray_face, -1)
emotion_prediction = emotion_classifier.predict(gray_face)
emotion_probability = np.max(emotion_prediction)
emotion_label_arg = np.argmax(emotion_prediction)
emotion_text = emotion_labels[emotion_label_arg]
# print 'the emotion probability is:',emotion_probability
# print 'the emotion is:',emotion_label_arg
# print 'the predicted emotion is:',emotion_text
return emotion_text
def image_generator(X_image, X_flip, y_steering, batch_size):
sample_image_shape = mpimg.imread(X_image[0]).shape
# Create empty arrays to contain batch of features and labels#
batch_features = np.zeros((batch_size, sample_image_shape[0], sample_image_shape[1], sample_image_shape[2]))
batch_labels = np.zeros((batch_size, 1))
while True:
for i in range(batch_size):
#choose random index in features
index = np.random.randint(len(X_image))
image = mpimg.imread(X_image[index])
if X_flip[index] == 0:
batch_features[i] = image
batch_labels[i] = y_steering[index]
else:
batch_features[i] = cv2.flip(image, flipCode=1)
batch_labels[i] = y_steering[index]
processed_feature = preprocessing.preprocess_input(batch_features)
yield [processed_feature, processed_feature], batch_labels
def telemetry(sid, data):
# The current steering angle of the car
steering_angle = data["steering_angle"]
# The current throttle of the car
throttle = data["throttle"]
# The current speed of the car
speed = data["speed"]
# The current image from the center camera of the car
imgString = data["image"]
image = Image.open(BytesIO(base64.b64decode(imgString)))
image_array = np.asarray(image)
transformed_image_array = image_array[None, :, :, :]
transformed_image_array = preprocessing.preprocess_input(
transformed_image_array)
# This model currently assumes that the features of the model are just the images. Feel free to change this.
steering_angle = float(
model.predict([transformed_image_array, transformed_image_array],
batch_size=1))
# The driving model currently just outputs a constant throttle. Feel free to edit this.
throttle = 0.2
print(steering_angle, throttle)
send_control(steering_angle, throttle)
def main():
# Read dataset
dict = index_dataset()
# Preprocess dataset if needed
if not os.path.exists('./objects/indexer.pickle') or not os.path.exists(
'./objects/knn.pickle'):
dataset, corpus = preprocess_dataset(dict,
lemmatize=True,
remove_stopwords=True,
measure_time=True)
# Load or create indexer
if os.path.exists('./objects/indexer.pickle'):
indexer = load_object('./objects/indexer.pickle')
else:
indexer = Indexer(dataset, measure_time=True)
save_object(indexer, './objects/indexer.pickle')
#Load or create KNN
if os.path.exists('./objects/knn.pickle'):
knn = load_object('./objects/knn.pickle')
else:
# Initialize KNN with given dataset
knn = KNN(dataset, corpus, measure_time=True)
save_object(knn, './objects/knn.pickle')
# Main loop for user input
print("Type a question:")
q = input()
while q != 'quit':
processed_input = preprocess_input(q,
lemmatize=True,
remove_stopwords=True)
terms_to_search_for = list(processed_input.keys())
print('Terms to search for:')
print(terms_to_search_for)
print()
containing_docs = indexer.retrieve_documents(terms_to_search_for,
measure_time=True)
res = knn.find_nearest_neigbours(processed_input,
containing_docs,
k=10,
measure_time=True)
print("\nResults:\n")
i = 1
for r in res:
print(f'#{i}')
print(r)
print()
i += 1
print("Type a question:")
q = input()
import cv2
import numpy as np
import os
import preprocessing as pre
# Getting required directories
# Model Directory
mod_dir = "/".join(os.getcwd().split("/")[0:-1] + ['model/'])
# Data Directory
data_dir = "/".join(os.getcwd().split("/")[0:-1] + ['data/'])
# Base Directory
base_dir = "/".join(os.getcwd().split("/")[0:-1])
# data loading and preprocessing
faces, emotions = pre.load_data(data_dir + "fer2013.csv")
faces = pre.preprocess_input(faces)
xtrain, xtest, ytrain, ytest = train_test_split(faces,
emotions,
test_size=0.2,
shuffle=True)
# parameters
batch_size = 64
num_epochs = 100
input_shape = (48, 48, 1)
verbose = 1
num_classes = 7
patience = 30
l2_regularization = 0.01
# data generator
def main():
input_file = os.path.abspath(args.input_file)
data_output_dir = os.path.abspath(args.data_output_dir)
final_output_dir = os.path.abspath(args.final_output_dir)
if not os.path.exists(data_output_dir):
os.makedirs(data_output_dir)
if not os.path.exists(final_output_dir):
os.makedirs(final_output_dir)
# Support some standard preprocessing
if args.tokenize:
# tokenize input_file to token_file
logging.info("tokenizing data")
token_file = "%s/tokens.jsonlist.gz" % data_output_dir
func = functools.partial(preprocessing.tokenize,
filter_stopwords=args.nostopwords)
preprocessing.preprocess_input(input_file, token_file)
input_file = token_file
if args.lemmatize:
# lemmatize input_file to lemma_file
logging.info("lemmatizing data")
lemma_file = "%s/lemmas.jsonlist.gz" % data_output_dir
func = functools.partial(preprocessing.lemmatize,
filter_stopwords=args.nostopwords)
preprocessing.preprocess_input(input_file, lemma_file)
input_file = lemma_file
# generate topics or lexicons
option = args.option
num_ideas = args.num_ideas
cooccur_func = functools.partial(il.generate_cooccurrence_from_int_set,
num_ideas=num_ideas)
prefix = args.prefix
if option == "topics":
logging.info("using topics to represent ideas")
prefix = "%s_topics" % prefix
# generate mallet topics
mt.get_mallet_input_from_words(input_file, data_output_dir)
if not mt.check_mallet_directory(data_output_dir):
# run mallet to prepare topics inputs
# users can also generate mallet-style topic inputs inputs
logging.info("running mallet to get topics")
if not os.path.exists(os.path.join(args.mallet_bin_dir, 'mallet')):
sys.exit("Error: Unable to find mallet at %s" %
args.mallet_bin_dir)
os.system("./mallet.sh %s %s %d" %
(args.mallet_bin_dir, data_output_dir, num_ideas))
# load mallet outputs
articles, vocab, idea_names = mt.load_articles(input_file,
data_output_dir)
table_top = 5
elif option == "keywords":
logging.info("using keywords to represent ideas")
prefix = "%s_keywords" % prefix
# idenfity keyword ideas using fighting lexicon
lexicon_file = "%s/fighting_lexicon.txt" % data_output_dir
other_files = [args.background_file]
fl.get_top_distinguishing(input_file, other_files, data_output_dir,
lexicon_file)
# load keywords
articles, word_set, idea_names = fl.load_word_articles(
input_file, lexicon_file, data_output_dir, vocab_size=num_ideas)
table_top = 10
else:
logging.error("unsupported idea representations")
# compute strength between pairs and generate outputs
il.generate_all_outputs(articles,
num_ideas,
idea_names,
prefix,
final_output_dir,
cooccur_func,
table_top=table_top,
group_by=args.group_by)
def predict_vm(ecs_lines, input_lines):
# Do your work from here#
result = []
if ecs_lines is None:
return result
if input_lines is None:
return result
mission = preprocessing.preprocess_input(input_lines)
flavor_dict = preprocessing.preprocess_ecs_info(ecs_lines, mission)
data_dict_merge = preprocessing.merge(flavor_dict, mission)
data_dict_filled = preprocessing.fill_data(data_dict_merge, mission)
error_point = ['']
if len(mission.vm_type) == 3:
if mission.opt_target == "CPU":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 53
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=405)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
elif mission.opt_target == "MEM":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 55
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=40)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
elif len(mission.vm_type) == 5:
if mission.opt_target == "CPU":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 46
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=405)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
elif mission.opt_target == "MEM":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 60
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=40)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
else:
if mission.opt_target == "CPU":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 42
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=405)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
elif mission.opt_target == "MEM":
flavname = data_dict_filled.keys()
nn_parameter_dict = dict()
deg = 48
for name in flavname:
data = data_dict_filled[name]
series = [x[1] for x in data]
X, Y = generate_samples(series, deg)
X = X.transpose()
Y = Y.transpose()
parameters = nn_model(X,
Y,
4,
learning_rate=0.01,
num_iterations=40)
nn_parameter_dict[name] = parameters
# 读取需要预测的天数
start_time = datetime.datetime.strptime(
mission.time_limit[0].split(' ')[0], '%Y-%m-%d')
end_time = datetime.datetime.strptime(
mission.time_limit[1].split(' ')[0], '%Y-%m-%d')
days = (end_time - start_time).days
pred_result_dict = dict()
pred_list = []
for name in flavname:
# 取数据的最后degree个样本点
n_sum = 0 # flavor在该周期内的数量总和
flavor_data = data_dict_filled[name] # 数据点
parameters = nn_parameter_dict[name] # 系数
samples = flavor_data[-deg:]
samples = [x[1] for x in samples]
samples = matrix([samples])
samples = samples.transpose()
temp = []
for day in range(days):
A2, cache = forward_propagation(samples, parameters)
x_pred = A2.value[0][0]
x_pred = math.ceil(abs(x_pred))
n_sum = n_sum + x_pred
samples.value[0][0] = x_pred
temp.append(x_pred)
pred_result_dict[name] = n_sum
print("预测完毕")
# 放置服务器
phy_server = annealingoptimize(pred_result_dict,
mission,
T=10000.0,
cool=0.98)
print("放置完成")
print("生成结果输出文件")
n_sum = 0
for name in pred_result_dict.keys():
n_sum = n_sum + pred_result_dict[name]
result.append(int(n_sum))
for name in pred_result_dict.keys():
result.append("%s %d" % (name, pred_result_dict[name]))
result.append("")
result.append(len(phy_server))
for n in phy_server.keys():
s = str(n)
u_name = set(phy_server[n])
for u_n in u_name:
cnt = phy_server[n].count(u_n)
s = s + ' ' + str(u_n) + ' ' + str(cnt)
result.append(s)
return result
def main(_):
if not os.path.exists("./outputs/steering_model"):
os.makedirs("./outputs/steering_model")
batch_size = 1024
X_image = []
X_flip = []
y_steering = []
with open(FLAGS.drive_log_file, 'rt') as csvfile:
reader = csv.reader(csvfile, skipinitialspace=True)
for row in reader:
if row[0] == 'center':
continue
X_image.append(row[0])
X_flip.append(0)
y_steering.append(float(row[3]))
X_image.append(row[1])
X_flip.append(0)
y_steering.append(float(row[3]) + 0.15)
X_image.append(row[2])
X_flip.append(0)
y_steering.append(float(row[3]) - 0.15)
# X_image.append(row[0])
# X_flip.append(0)
# y_steering.append(float(row[3]))
X_image.append(row[1])
X_flip.append(1)
y_steering.append(float(row[3]) - 0.15)
X_image.append(row[2])
X_flip.append(1)
y_steering.append(float(row[3]) + 0.15)
X_image = np.array(X_image)
X_flip = np.array(X_flip)
y_steering = np.array(y_steering)
X_train_image, X_test_image, X_train_flip, X_test_flip, y_train_steering, y_test_steering = train_test_split(
X_image, X_flip, y_steering, test_size=0.2)
if len(X_train_image) == 0:
print('No data found')
return
sample_image = preprocessing.preprocess_input(
np.array([mpimg.imread(X_train_image[0])]))[0]
print('Image Shape: ', sample_image.shape)
sample_per_epoch = len(X_train_image)
nb_val_samples = len(X_test_image)
print('Training Data Count:', sample_per_epoch)
print('Validation Data Count:', nb_val_samples)
X_train_image, X_train_flip, y_train_steering = shuffle(
X_train_image, X_train_flip, y_train_steering)
comma_ai_model = get_comma_ai_model(sample_image.shape)
# if there is a model given, then we will train on it, else we will train on a new model
if FLAGS.model_file != '':
model = load_model(FLAGS.model_file)
else:
model = comma_ai_model
model.compile(optimizer="adam", loss="mse")
model.fit_generator(image_generator(X_train_image, X_train_flip,
y_train_steering, batch_size),
samples_per_epoch=sample_per_epoch,
nb_epoch=1,
validation_data=image_generator(
X_test_image, X_test_flip, y_test_steering,
batch_size),
nb_val_samples=nb_val_samples)
print("Saving model weights and configuration file.")
model.save("./outputs/steering_model/steering_angle.h5", True)