def datadealing(City):
# print u"收到请求", request.path, request.args
# a = time.clock()
Mode = request.args["Mode"]
k = dataProcess(Mode, City)
# print u"用了时间", time.clock() - a
return jsonify(k if Mode != u"rawJson" else {u"rawJson": k})
def load_data(self, npy_dir, is_train=True):
mydata = dataProcess(self.img_rows, self.img_cols)
if (is_train):
img_train, label_train = mydata.load_train_data(npy_dir)
return img_train, label_train
else:
img_test, label_test = mydata.load_test_data(npy_dir)
return img_test, label_test
def load_train_data(self): mydata = dataProcess(self.img_rows, self.img_cols) imgs_train, imgs_mask_train = mydata.load_train_data() # preprocessing imgs_train = DataProcess.scale(imgs_train) imgs_mask_train = DataProcess.scale(imgs_mask_train, flag_mask=True) return imgs_train, imgs_mask_train
def load_train_data(self): mydata = dataProcess() imgs_train, mask_train = mydata.load_train_data() # preprocessing imgs_train = DataProcess.preprocess(imgs_train, self.img_rows, self.img_cols) mask_train = DataProcess.preprocess(mask_train, self.img_rows, self.img_cols, flag_mask=True) return imgs_train, mask_train
def predict(self, image, model):
mydata = dataProcess(self.img_rows, self.img_cols)
imgs_test = mydata.creat_and_load_single_test_image_data(image)
print("loading data done")
print('predict test data')
imgs_mask_test = model.predict(imgs_test, batch_size=1, verbose=1)
imgs = imgs_mask_test
img = imgs[0] #因为只有一张图像
img = array_to_img(img)
return img
def addLocation(City, StationName, MarsLat, MarsLon):
a = dataProcess(City=City, Mode="allStations")
for station in a[u"Stations"]:
if station["Name"] == StationName:
station["MarsLat"] = MarsLat
station["MarsLon"] = MarsLon
file = open(u"./cache/{City}_allStations.json".format(
City=City), "w")
file.write(json.dumps(a))
file.close()
return getStationPP(City, StationName)
def predict(self, model):
print('load test data')
# load data
mydata = dataProcess(self.img_rows, self.img_cols)
imgs_test = mydata.load_test_data()
# preprocess
imgs_test = DataProcess.scale(imgs_test)
print('predict test data')
imgs_mask_test = model.predict(imgs_test, batch_size=1, verbose=1)
np.save(self.test_result_name(), imgs_mask_test)
print('save results to jpeg files')
self.save_img()
def load_model_and_predict(self, st_num, end_num):
model = self.get_unet()
# 加载保存的模型权重
model.load_weights('unet.hdf5')
for i in range(st_num, end_num):
mydata = dataProcess(self.img_rows, self.img_cols)
mydata.create_single_test_data(i)
# mydata.create_test_data()
imgs_test = self.load_test_data()
print("loading data done")
print('predict test data')
imgs_mask_test = model.predict(imgs_test, batch_size=1, verbose=1)
np.save('../results/imgs_mask_test.npy', imgs_mask_test)
self.save_single_img(i)
def use_saved_model(self, num):
mydata = dataProcess(self.img_rows, self.img_cols)
mydata.create_single_test_data(num)
#mydata.create_test_data()
imgs_test = self.load_test_data()
print("loading data done")
model = self.get_unet()
#加载保存的模型权重
model.load_weights('unet.hdf5')
# compile 编译
#model.compile(metrics=['accuracy'])
print('predict test data')
imgs_mask_test = model.predict(imgs_test, batch_size=1, verbose=1)
np.save('../results/imgs_mask_test.npy', imgs_mask_test)
def load_test_data(self):
mydata = dataProcess(self.img_rows, self.img_cols)
imgs_test = mydata.load_test_data()
return imgs_test
def load_data(self):
mydata = dataProcess(self.img_rows, self.img_cols)
imgs_train, imgs_mask_train = mydata.load_train_data()
imgs_test = mydata.load_test_data()
return imgs_train, imgs_mask_train, imgs_test
activation='relu')(model)
model = concatenate([model_conn4, model_tr], axis=3)
model = Conv2D(32, (3, 3), activation='relu', padding='same')(model)
model_tr = Conv2DTranspose(32, (3, 3),
strides=2,
padding='same',
activation='relu')(model)
model = concatenate([model_conn5, model_tr], axis=3)
model = Conv2D(32, (3, 3), activation='relu', padding='same')(model)
model = Conv2D(1, (1, 1), activation='sigmoid', padding='same')(model)
instru = Model(input=inputs, output=model)
instru.compile(optimizer=Adam(lr=1e-4), loss=IOU_calc_loss, metrics=[IOU_calc])
mydata = dataProcess(1024, 1280)
imgs_train, imgs_mask_train = mydata.load_train_data()
print(imgs_mask_train.shape)
print('Loading done')
model_checkpoint = ModelCheckpoint('../drive/checkpoints/instru.hdf5',
monitor='loss',
verbose=1,
save_best_only=True)
instru.load_weights('../drive/checkpoints/instru.hdf5')
instru.fit(imgs_train,
imgs_mask_train,
batch_size=2,
nb_epoch=3,
verbose=1,
validation_split=0.2,
shuffle=True,
from keras.callbacks import ModelCheckpoint
import data, model
dataproc = data.dataProcess(299, 299)
dataproc.create_train_data()
imgs, mask = dataproc.load_train_data()
model = model.get_model()
model_checkpoint = ModelCheckpoint('model.{epoch:02d}-{val_loss:.2f}.hdf5',
verbose=0,
period=50)
print('Fitting model...')
model.fit(imgs,
mask,
batch_size=4,
epochs=10000,
verbose=1,
validation_split=0.2,
shuffle=True,
callbacks=[model_checkpoint])
def City(City):
if "Attention" in dataProcess("rawJson", City):
for attention in dataProcess("rawJson", City)["Attention"]:
flash(attention, "warning")
return render_template('city.html', lista=List, City=City)
def load_test_data(self):
mydata = dataProcess(self.img_rows, self.img_cols, mode=self.mode)
imgs_test, imgs_label_test = mydata.load_test_data()
return imgs_test, imgs_label_test
def load_data(self): mydata = dataProcess(self.img_rows, self.img_cols) imgs_train, imgs_mask_train = mydata.load_train_data() imgs_test = mydata.load_test_data() return imgs_train, imgs_mask_train, imgs_test
