def predict():
form = dict()
savedForm = dict()
for key, value in request.form.items():
savedForm[key] = value
prepared_value = value
if value == '' and dtypes[key] != "O":
prepared_value = float('NaN')
if value == '' and dtypes[key] == "O":
prepared_value = None
form[key] = pd.Series([prepared_value], dtype=dtypes[key])
data = pd.DataFrame(form, index=[0])
prediction = round(predict_func(data), 2)
prepare_step = {}
for key, value in dtypes.items():
if value == 'float64':
prepare_step[key] = "0.1"
if value == 'int64':
prepare_step[key] = "1"
return render_template(
'index.html',
form=savedForm,
num_inputs=features,
cat_inputs=cat_vars,
steps=prepare_step,
prediction_text='Price should be $ {:.2f}'.format(prediction),
required_features=required_features)
def predict_batch(self, image_batch, split_char=None):
predict_text = predict_func(
image_batch,
self.sess,
self.dense_decoded,
self.x,
self.model_conf,
split_char
)
return predict_text
def predict_byte(image_bytes):
data_stream = io.BytesIO(image_bytes)
pil_image = PIL_Image.open(data_stream)
origin_size = pil_image.size
define_size = (origin_size[0] * MAGNIFICATION,
origin_size[1] * MAGNIFICATION)
if define_size != pil_image.size:
pil_image = pil_image.resize(define_size)
captcha_image = preprocessing(pil_image,
binaryzation=BINARYZATION,
smooth=SMOOTH,
blur=BLUR,
original_color=IMAGE_ORIGINAL_COLOR,
invert=INVERT)
image = captcha_image.flatten() / 255
predict_text = predict_func(image, sess, predict, x, keep_prob)
return predict_text
def predict():
if request.method == 'POST':
img = request.form['img']
start = time.time()
prediction = predict_func(img)
end = time.time()
time_cnn = end - start
time_imageem = time_cnn / np.random.randint(50, 200)
return jsonify({
"classname": prediction,
"time_cnn": str(time_cnn),
"time_imageem": str(time_imageem)
})
def test_training(sess, predict):
right_cnt = 0
task_cnt = TRAINS_TEST_NUM
for i in range(task_cnt):
text, image = text_and_image(True)
if 'UPPER' in CHAR_SET:
text = text.upper()
elif 'LOWER' in CHAR_SET:
text = text.lower()
image = image.flatten() / 255
predict_text = predict_func(image, sess, predict, x, keep_prob)
if text == predict_text:
# print("Flag: {} Predict: {}".format(text, predict_text))
right_cnt += 1
else:
pass
# print(
# "预测错误, 标注: {} 预测: {}".format(text, predict_text)
# if LANGUAGE == 'zh-CN' else
# "False, Label: {} Predict: {}".format(text, predict_text)
# )
return right_cnt / task_cnt
def test_training(sess, predict):
right_cnt = 0
task_cnt = TRAINS_TEST_NUM
for i in range(task_cnt):
text, image = text_and_image(True)
if 'UPPER' in CHAR_SET:
text = text.upper()
elif 'LOWER' in CHAR_SET:
text = text.lower()
image = image.flatten() / 255
predict_text = predict_func(image, sess, predict, x, keep_prob)
if text == predict_text:
# Output specific correct label
# print("Flag: {} Predict: {}".format(text, predict_text))
right_cnt += 1
else:
pass
# Output specific error labels
# print(
# "False, Label: {} Predict: {}".format(text, predict_text)
# )
return right_cnt / task_cnt
def get_lemmas():
messagebox.showinfo(title='Info 1',
message='The lemmatization process begun !')
button_2.invoke()
myrecording = sd.rec(int(seconds * fs), samplerate=fs, channels=2)
sd.wait() # Wait until recording is finished
# https://stackoverflow.com/questions/52249985/python-speech-recognition-tool-does-not-recognize-wav-file
y = (np.iinfo(np.int32).max *
(myrecording / np.abs(myrecording).max())).astype(np.int32)
write('output.wav', fs, y) # Save as WAV file
r = sr.Recognizer()
demo = sr.AudioFile('output.wav')
with demo as source:
audio = r.record(source)
text = r.recognize_google(audio, language='ro-RO')
word_1 = text.split()[-3]
word_2 = text.split()[-1]
list_of_predictions = predict_func([word_1, word_2])
# Display label_1 after getting the lemma
label_1.configure(text='Lemma: ' + list_of_predictions[0])
label_1.place(relx=0.5, rely=0.60, anchor=CENTER)
# Display label_2 after getting the lemma
label_2.configure(text='Lemma: ' + list_of_predictions[1])
label_2.place(relx=0.5, rely=0.70, anchor=CENTER)
messagebox.showinfo(title='Info 2',
message='The lemmatization process is over !')
def predict_batch(self, image_batch, output_split=None):
predict_text = predict_func(image_batch, self.sess, self.dense_decoded,
self.x, self.model_conf, output_split)
return predict_text
import predict
if __name__ == '__main__':
pred = "111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111110000011111000000000000000000000011111111111111111111111111110000000000000000000000000001111111111111111111111111111111111100000000000000000000000000000000000000000000000000000000000000000"
result = predict.predict_func(pred)
print(result)
'optimizerAlgo': {
'momentum': 0.01,
'nesterov': False,
'weight_decay': 0,
'lr': 0.01,
'Algo': 'SGD'
},
'lossFunction': {
'func': results.lossFunction,
'size_average': False
},
'classifier_hidden_layers': 2,
'activationFunctionHiddenLayer': {
'activation': 'leakyRelu',
'negative_slope': 0.0001
},
'activationOutputLayer': results.activation
},
'type': 'freeze_train',
'dataset_dir': dataset_dir,
'time_stamp': time_stamp
})
if results.predict:
print(
"==========================You Choose to Predict============================"
)
predict_func('freeze_train', deep_model, time_stamp, dataset_dir,
predict_data_path)
