def compute_saliency(model,
guided_model,
img_path,
preprocessed_input,
layer_name='block5_conv3',
cls=-1,
visualize=True,
save=True):
"""Compute saliency using all three approaches.
-layer_name: layer to compute gradients;
-cls: class number to localize (-1 for most probable class).
"""
#preprocessed_input = load_image(img_path)
predictions = model.predict(preprocessed_input)
top_n = 5
top = decode_predictions(predictions, top=top_n)[0]
classes = np.argsort(predictions[0])[-top_n:][::-1]
print('Model prediction:')
for c, p in zip(classes, top):
print('\t{:15s}\t({})\twith probability {:.3f}'.format(p[1], c, p[2]))
if cls == -1:
cls = np.argmax(predictions)
class_name = decode_predictions(np.eye(1, 1000, cls))[0][0][1]
print("Explanation for '{}'".format(class_name))
gradcam = grad_cam(model, preprocessed_input, cls, layer_name)
gb = guided_backprop(guided_model, preprocessed_input, layer_name)
guided_gradcam = gb * gradcam[..., np.newaxis]
if save:
jetcam = cv2.applyColorMap(np.uint8(255 * gradcam), cv2.COLORMAP_JET)
jetcam = (np.float32(jetcam) +
load_image(img_path, H, W, preprocess=False)) / 2
cv2.imwrite('gradcam.jpg', np.uint8(jetcam))
cv2.imwrite('guided_backprop.jpg', deprocess_image(gb[0]))
cv2.imwrite('guided_gradcam.jpg', deprocess_image(guided_gradcam[0]))
if visualize:
plt.figure(figsize=(15, 10))
plt.subplot(131)
plt.title('GradCAM')
plt.axis('off')
plt.imshow(load_image(img_path, H, W, preprocess=False))
plt.imshow(gradcam, cmap='jet', alpha=0.5)
plt.subplot(132)
plt.title('Guided Backprop')
plt.axis('off')
plt.imshow(np.flip(deprocess_image(gb[0]), -1))
plt.subplot(133)
plt.title('Guided GradCAM')
plt.axis('off')
plt.imshow(np.flip(deprocess_image(guided_gradcam[0]), -1))
plt.show()
return gradcam, gb, guided_gradcam
def confront_prediction(file):
#加载图片并进行处理，使之符合网络的输入格式
img = image.load_img(file, target_size=(299, 299)) #加载猫的照片
input_image = image.img_to_array(img) #将其转换为数组(299,299,3)
# 归一化图片到[-1,1.]
input_image /= 255. #变为0-1
input_image -= 0.5 #变为-0.5—0.5
input_image *= 2. #变为-1—1
# 为图片添加第四维（batch），使之符合模型的输入
input_image = np.expand_dims(
input_image,
axis=0) #用于扩展数组的形状，axis=0表示在0的地方扩展形状，将(299,299,3)变为（1，299，299，3）
# model = inception_v3.InceptionV3(weights='imagenet')#从网络加载
# model = load_model("static/trained_models/inception_v3_on_imageNet.h5")#从本地加载
#进行预测，并输出预测结果
global graph
with graph.as_default():
predictions = model.predict(input_image) #使用模型进行预测
predicted_classes = inception_v3.decode_predictions(
predictions, top=1) #top=1表示输出概率最高的前1个类别
# print(predicted_classes)#查看输出的结果
imagenet_id, name, confidence = predicted_classes[0][0] #取概率最高的预测类别的信息
# print("This is a {0} with {1:.4}% confidence!".format(name, confidence * 100))
return "This is a {0}!".format(name)
def predict_from_saved_model(model_path, image_path, decode_fn, top=5):
predict_fn = predictor.from_saved_model(export_dir=model_path)
with open(image_path, 'rb') as f:
b64_x = f.read()
b64_x = base64.urlsafe_b64encode(b64_x)
input_instance = {'inputs': [b64_x]}
preds = predict_fn(input_instance)['outputs'][0]
if decode_fn == 'inception':
preds = np.expand_dims(preds, 0)
print('Predicted:')
for p in decode_predictions(preds, top=5)[0]:
print("Score {}, Label {}".format(p[2], p[1]))
elif decode_fn == 'transfert':
preds = [(fruit_mapping[idx], pred) for idx, pred in enumerate(preds)]
preds = sorted(preds, key=lambda pred: pred[1], reverse=True)
if top:
preds = preds[:top]
print('Predicted:')
for pred in preds:
print("Score {}, Label {}".format(pred[1], pred[0]))
else:
print('Predicted scores:')
print(preds)
def predict(image_file):
img = image.load_img(image_file, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
global graph
with graph.as_default():
preds = model.predict(x)
top3 = decode_predictions(preds, top=3)[0]
answer = dict(zip(('label', 'description', 'probability'), top3[0]))
print(answer)
predictions = [{
'label': label,
'description': description,
'probability': round(probability, 2)
} for label, description, probability in top3]
return {
'answer': {
'label': answer['label'],
'description': answer['description'],
'probability': round(answer['probability'], 2)
},
'top3': predictions
}
def imagelabel2json(tmpfile, model):
print("画像解析開始")
img = image.load_img(tmpfile, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
recognize = decode_predictions(preds)
en_label = recognize[0][0][1]
label = en_label
score = str(recognize[0][0][2])
with open(IMAGENET_JSON_PATH, 'r') as f:
obj = json.load(f)
for i in obj:
if i['en'] == en_label:
label = i['ja']
break
dict = {'label': label, 'score': score}
f = open(LABEL_JSON_PATH, 'w')
json.dump(dict, f)
print("画像解析終了")
def classify(img):
arr = preprocess_input(np.expand_dims(img.astype(np.float32), axis=0))
preds = model.predict(arr)
_, lab, p = decode_predictions(preds, top=3)[0][0]
if p < 0.5:
lab = ''
return lab
def identifica(self, nomeImagem):
""" [Identificação de imagem por rede neural]
Arguments:
nomeImagem {[string]} -- [nome da imagem a ser identificada pela rede neural]
Returns:
[Lista] -- [Contendo as 5 identificaçoes encontrada da possibilidade
de reconhecimento da imagem recebida.]
"""
# Rede neural identificando imagem.
ar_X = image.img_to_array(image.load_img(nomeImagem, target_size=(299, 299)))
# Alterando os valores da matrix de um ranger de 0-255 para -1 a 1.
ar_X /= 255
ar_X -= 0.5
ar_X *= 2
# aplicando reshape.
ar_X = ar_X.reshape([1, ar_X.shape[0], ar_X.shape[1], ar_X.shape[2]])
ar_Y = self.iv3.predict(ar_X)
# apos processamento retorna a classificação processada.
resultado = decode_predictions(ar_Y)
self.X = np.copy(ar_X)
return resultado
def predict(model, img_names, target_size, top_n=3):
"""Run model prediction on image
Args:
model: keras model
img: PIL format image
target_size: (w,h) tuple
top_n: # of top predictions to return
Returns:
list of predicted labels and their probabilities
"""
xs = []
for img_name in img_names:
img = Image.open(img_name)
if img.size != target_size:
img = img.resize(target_size)
x = image.img_to_array(img)
xs.append(x)
xs = np.asarray(xs)
xs = preprocess_input(xs)
preds = model.predict(xs)
res = decode_predictions(preds, top=top_n)
for i in range(len(img_names)):
print img_names[i]
print res[i]
print
def main():
base_model = load_base_model('ResNet50', None)
# Use correct image preprocessing for model
if base_model.name == ('inception_v3'):
preprocess_input = inception_v3_preprocess_input
else:
preprocess_input = preprocess_input_wrapper
# Rewrite this bad boy
waitForIn = threading.Thread(target=waitForTerminate)
waitForIn.start()
while True:
images, filenames = load_images('image', 224)
if len(filenames) == 0:
if kill:
print("The user terminated the program\n")
exit(0)
time.sleep(2)
else:
if kill:
print("The user terminated the program\n")
exit(0)
preds = base_model.predict(images)
print()
print(filenames[0])
print()
print('Predicted:', decode_predictions(preds, top=10))
os.remove('image/' + filenames[0])
def predict(self, model, image_path):
'''
inputs:
image_paths: list of image paths or can be a string too (for single image)
'''
image = load_img(image_path, target_size=self.image_size)
image = img_to_array(image)
image = image.reshape(
(1, image.shape[0], image.shape[1], image.shape[2]))
image = preprocess_input(image)
model_preds = model.predict(image)
sort_index_preds = numpy.flip(numpy.argsort(model_preds))[0]
accumulate = 0.
res = []
if self.model == 'InceptionV3':
pr = decode_predictions(model_preds, top=int(self.max_items))[0]
accumulate = 0.
for i in pr:
if i[2] < self.min_threshold:
return res, accumulate
accumulate += i[2]
res.append(i[1])
if accumulate > self.threshold:
return res, accumulate
return res, accumulate
for i in sort_index_preds:
if model_preds[0][i] < self.min_threshold:
return res, accumulate
accumulate += model_preds[0][i]
res.append(self.categories[i])
if accumulate > self.threshold:
return res, accumulate
def my_detection(filename):
# Load pre-trained image recognition model
model = inception_v3.InceptionV3()
# Load the image file and convert it to a numpy array
img = image.load_img(filename, target_size=(299, 299))
input_image = image.img_to_array(img)
print(input_image.shape)
# Scale the image so all pixel intensities are between [-1, 1] as the model expects
input_image /= 255.
input_image -= 0.5
input_image *= 2.
# Add a 4th dimension for batch size (as Keras expects)
input_image = np.expand_dims(input_image, axis=0)
# Run the image through the neural network
print(input_image.shape)
predictions = model.predict(input_image)
# Convert the predictions into text and print them
predicted_classes = inception_v3.decode_predictions(predictions, top=1)
imagenet_id, name, confidence = predicted_classes[0][0]
print("This is a {} with {:.4}% confidence!".format(
name, confidence * 100))
def open(self):
for ele in root.winfo_children():
ele.destroy()
myFormats = [('JPEG / JFIF','*.jpg')]
global file_path_string
file_path_string = tkFileDialog.askopenfilename(filetypes=myFormats)
im = Image.open(file_path_string)
tkimage = ImageTk.PhotoImage(im)
myvar = Label(root,image = tkimage)
myvar.image = tkimage
myvar.pack()
global file_path_string
img_path = file_path_string
img = image.load_img(img_path, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
global model
preds = model.predict(x)
global file_path_string
myvar = Label(root, text=file_path_string + "\n" + str(decode_predictions(preds, top=3)[0]))
myvar.pack()
self.open_button = Button(root, text="Open File", command=self.open)
self.open_button.pack()
root.close_button = Button(root, text="Close Program", command=root.quit)
root.close_button.pack()
def predict():
# initialize the data dictionary that will be returned from the
# view
data = {"success": False}
# ensure an image was properly uploaded to our endpoint
if flask.request.method == "POST":
if flask.request.files.get("image"):
# read the image in PIL format
image = flask.request.files["image"].read()
image = Image.open(io.BytesIO(image))
# preprocess the image and prepare it for classification
image = prepare_image(image, target=(224, 224))
# classify the input image and then initialize the list
# of predictions to return to the client
preds = model.predict(image)
results = decode_predictions(preds)
data["predictions"] = []
# loop over the results and add them to the list of
# returned predictions
for (imagenetID, label, prob) in results[0]:
r = {"label": label, "probability": float(prob)}
data["predictions"].append(r)
# indicate that the request was a success
data["success"] = True
# return the data dictionary as a JSON response
return flask.jsonify(data)
def get(self, resource_id):
arr = [
'taj.jpg', 'state.jpg', 'yos.jpg', 'waipio.jpg', 'owens.jpg',
'sonoma.jpg'
]
model = tf.keras.applications.inception_v3.InceptionV3(
include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000)
# model = tf.keras.models.load_model('./inception.h5')
for xi in arr:
path = os.path.abspath('files/' + xi)
img = image.load_img(path, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
# print(preds)
print('Predicted:', decode_predictions(preds, top=5)[0])
return {'images': 'data'}
def classify(self, img_path, top=5):
"""Classify image and return top matches.
:param img_path: input file path of image.
:param top: number of top results to return.
:returns: predictions about detected classes in image.
:rtype: list[list[tuple(str: class_id, str: class_name, float: score)]]
"""
# Open image
img = Image.open(img_path)
# Image resizing and preparation for keras.
if img.size != self.target_size:
img = img.resize(self.target_size)
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = inception_v3.preprocess_input(x)
# Predictions
preds = []
with self.graph.as_default():
preds = self.inception_model.predict(x)
# Decode predictions
return inception_v3.decode_predictions(preds, top=top)
def interception(image:InceptionImageModel):
iv3 = InceptionV3()
URL = 'http://localhost:8000/images/' + str(image.image)
IMAGE_PATH = './media/temp/temp.png'
with urllib.request.urlopen(URL) as url:
with open(IMAGE_PATH, 'wb') as f:
f.write(url.read())
img = Image.open(IMAGE_PATH)
img = img.resize([299, 299])
x = keras_image.img_to_array(img)
#cambio de rando
x /=255
x -= 0.5
x *=2
#redimencion
x = x.reshape(1 , x.shape[0] , x.shape[1] , x.shape[2])
x.shape
y = iv3.predict(x)
predict = decode_predictions(y)[0][0]
name = predict[1]
percentage = predict[2]
return name , percentage
def predict(model, img):
img = img.resize((299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
return decode_predictions(preds, top=3)[0]
def pretrained():
keras.applications.resnet50.ResNet50(include_top=True, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000)
model = ResNet50(weights='imagenet')
'''
img_path = 'chair1.jpg'
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
'''
images = glob.glob('p2a-c/*.jpg')
i = 0
for fname in images:
img = image.load_img(fname, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print "-------------------"
print "FILE NAME: " + str(i)
print fname
print "-------------------"
print ""
print ""
i = i + 1
preds = model.predict(x)
print('Predicted:', decode_predictions(preds, top=3)[0])
print ""
print ""
print ""
print "See Confusion matrix in attached .xslx sheet"
print ""
print ""
print ""
print "Accuracy is 0.7586"
print ""
print ""
print ""
print "Recall is 0.7545"
print ""
print ""
print ""
print "Precision is 0.9152"
print ""
print ""
print ""
print "The f-score is 0.857"
def on_message(client, userdata, msg):
# mqtt_message = json.dumps(
# {
# "hash": hash,
# "image_location": dst,
# "color": color,
# "robot": "resource:com.diy3.Robot#1152",
# "asset": "resource:com.diy3.CapturedImage#"+str(ciid)})
matches = msg.payload.decode("utf-8")
matches = ast.literal_eval(matches)
print(matches['color'])
image_location = matches['image_location']
asset = matches['asset']
print(matches['image_location'])
print(matches['robot'])
print(matches['hash'])
# think I am going to need to call classify_image.py
img = image.load_img(image_location, target_size=(299, 299))
input_image = image.img_to_array(img)
# Scale the image so all pixel intensities are between [-1, 1] as the model expects
input_image /= 255.
input_image -= 0.5
input_image *= 2.
# Add a 4th dimension for batch size (as Keras expects)
input_image = np.expand_dims(input_image, axis=0)
# Run the image through the neural network
predictions = model.predict(input_image)
# Convert the predictions into text and print them
predicted_classes = inception_v3.decode_predictions(predictions, top=1)
imagenet_id, name, confidence = predicted_classes[0][0]
print("This is a {} with {:.4}% confidence!".format(
name, confidence * 100))
nameLikelihood = confidence * 100
nameLikelihood = str(round(nameLikelihood, 2)) + "%"
objectName = name.encode('utf-8')
# send to composer-rest-server api
time.sleep(5)
url = "http://<URL>:3000/api/com.diy3.ClassifyTransaction"
payload = "{\n \"$class\": \"com.diy3.ClassifyTransaction\",\n \"asset\": " '"' + asset + '"' ",\n \"ai\": \"resource:com.diy3.AI#9230\",\n \"nameLikelihood\": " '"' + nameLikelihood + '"' ",\n \"objectName\":" '"' + objectName + '"' " \n }"
# payload = "{\n \"$class\": \"com.diy3.ClassifyTransaction\",\n \"asset\":\""'"'+asset+'"'",\n \"ai\": \"resource:com.diy3.AI#9230\",\n \"nameLikelihood\": "'"'+nameLikelihood+'"'",\n \"objectName\": \""'"'+str(objectName)+'"'" \n}"
headers = {
'Content-Type': "application/json",
'Cache-Control': "no-cache",
}
response = requests.request("POST", url, data=payload, headers=headers)
print(response.text)
# completed api call
print('send transaction to composer-rest-server')
print('ready for the next image')
def predict(input_path):
model = inception_v3.InceptionV3()
formatted_image = image_formatter(input_path)
predictions = model.predict(formatted_image)
predicted_classes = inception_v3.decode_predictions(predictions)
imagenet_id, name, confidence = predicted_classes[0][0]
return name, confidence
def predict(self, path, top=5):
img = image.load_img(path, target_size=(self.size, self.size))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
pred = self._model.predict(x)
pred = decode_predictions(pred, top=top)[0]
return pred
def predict(self, img_path):
img = image.load_img(img_path, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = self.model.predict(x)
decoded = decode_predictions(preds)[0][0]
res = {'prediction': decoded[1], 'confidence': decoded[2]}
return res
def predict(image):
model = InceptionV3()
pred = model.predict(image)
decoded_predictions = decode_predictions(pred, top=10)
response = 'InceptionV3 predictions: ' + str(decoded_predictions[0][0:5])
print(response)
np.argmax(pred[0])
return response
def describe(model, input_image, verbose=False):
input_image = np.expand_dims(input_image, 0)
predictions = model.predict(input_image)
if verbose:
predicted_classes = decode_predictions(predictions, top=1)
imagenet_id, name, confidence = predicted_classes[0][0]
print("This is a {} with {:.4}% confidence!".format(name, confidence * 100))
return np.argmax(predictions)
def classify_inception(image_path):
try:
img = image.load_img(image_path, target_size=(224, 224))
except (OSError, IOError):
return [0, 0, 0.5]
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = inception_v3.preprocess_input(x)
preds = inception_model.predict(x)
return inception_v3.decode_predictions(preds, top=1)[0][0]
def classify(img):
h, w, _ = img.shape
dw = (w - 299) // 2
dh = (h - 299) // 2
win = img[dh:299 + dh, dw:299 + dw]
arr = preprocess_input(np.expand_dims(win.astype(np.float32), axis=0))
preds = model.predict(arr)
_, lab, p = decode_predictions(preds, top=3)[0][0]
if p < 0.5: lab = ''
return win, lab
def transformOutputToComparables(self, collected, output_col, get_uri):
values = {}
topK = {}
for row in collected:
uri = get_uri(row)
predictions = row[output_col]
self.assertEqual(len(predictions), ImageNetConstants.NUM_CLASSES)
values[uri] = np.expand_dims(predictions, axis=0)
topK[uri] = decode_predictions(values[uri], top=5)[0]
return values, topK
def transformOutputToComparables(self, collected, output_col, get_uri):
values = {}
topK = {}
for row in collected:
uri = get_uri(row)
predictions = row[output_col]
self.assertEqual(len(predictions), ImageNetConstants.NUM_CLASSES)
values[uri] = np.expand_dims(predictions, axis=0)
topK[uri] = decode_predictions(values[uri], top=5)[0]
return values, topK
def translateObject():
print(request.headers)
print(request.files['image1'])
image = request.files['image1']
image.save("object.png")
language = request.form['language']
print(language)
basewidth = 300
try:
image = Image.open("object.png")
for orientation in ExifTags.TAGS.keys():
if ExifTags.TAGS[orientation] == 'Orientation':
break
exif = dict(image._getexif().items())
if exif[orientation] == 3:
image = image.rotate(180, expand=True)
elif exif[orientation] == 6:
image = image.rotate(270, expand=True)
elif exif[orientation] == 8:
image = image.rotate(90, expand=True)
wpercent = (basewidth / float(image.size[0]))
hsize = int((float(image.size[1]) * float(wpercent)))
image = image.resize((basewidth, hsize), Image.ANTIALIAS)
image.save("object.png")
image.close()
except (AttributeError, KeyError, IndexError):
# cases: image don't have getexif
pass
target_size = (299, 299)
image = processImage("object.png", target_size)
with graph.as_default():
results = model.predict(image)
labels = decode_predictions(results)
print(labels)
translations = get_translation(labels[0], language)
print(translations)
print(labels)
response = []
for i in range(len(labels[0])):
translation = translations[i]
english = labels[0][i][1]
english = english.split("_")
english = " ".join(english)
print(english)
confidence = round(labels[0][i][2] * 100, 2)
response.append((translation, english, confidence))
json = {'predictions': response}
print(json)
return jsonify(json)
def _infer_classes(files_classes_proba: dict):
files_classes = {}
for fname, classes_proba in files_classes_proba.items():
proba_arr = np.expand_dims(np.array(classes_proba), 0)
class_pred = inception_v3.decode_predictions(
proba_arr, top=1)[0][0] # 1 for batch, 1 for top-n
class_name = class_pred[1] # we drop id and probability
class_name = _shorten_class_name(class_name)
files_classes[fname] = class_name
return files_classes
def classify_inception(image_path):
"""Classify image and return top match."""
img = Image.open(image_path)
target_size = (224, 224)
if img.size != target_size:
img = img.resize(target_size)
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = inception_v3.preprocess_input(x)
preds = inception_model.predict(x)
return inception_v3.decode_predictions(preds, top=1)[0][0]
def predict(image_file):
img = image.load_img(image_file, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x,axis=0)
x = preprocess_input(x)
global graph
with graph.as_default():
preds = model.predict(x)
top3 = decode_predictions(preds,top=3)[0]
predictions = [{'label': label, 'description': description, 'probability': probability * 100.0}
for label,description, probability in top3]
return predictions
def _executeTensorflow(self, graph, input_tensor_name, output_tensor_name,
df, id_col="filePath", input_col="image"):
with tf.Session(graph=graph) as sess:
output_tensor = graph.get_tensor_by_name(output_tensor_name)
image_collected = df.collect()
values = {}
topK = {}
for img_row in image_collected:
image = np.expand_dims(imageStructToArray(img_row[input_col]), axis=0)
uri = img_row[id_col]
output = sess.run([output_tensor],
feed_dict={
graph.get_tensor_by_name(input_tensor_name): image
})
values[uri] = np.array(output[0])
topK[uri] = decode_predictions(values[uri], top=5)[0]
return values, topK
def executeKerasInceptionV3(image_df, uri_col="filePath"):
"""
Apply Keras InceptionV3 Model on input DataFrame.
:param image_df: Dataset. contains a column (uri_col) for where the image file lives.
:param uri_col: str. name of the column indicating where each row's image file lives.
:return: ({str => np.array[float]}, {str => (str, str, float)}).
image file uri to prediction probability array,
image file uri to top K predictions (class id, class description, probability).
"""
K.set_learning_phase(0)
model = InceptionV3(weights="imagenet")
values = {}
topK = {}
for row in image_df.select(uri_col).collect():
raw_uri = row[uri_col]
image = loadAndPreprocessKerasInceptionV3(raw_uri)
values[raw_uri] = model.predict(image)
topK[raw_uri] = decode_predictions(values[raw_uri], top=5)[0]
return values, topK
