def run_batch():
print("Starting run...")
df = read_data()
data = df.to_dict(orient="records")
results = []
start = datetime.datetime.now()
for rows in np.array_split(data, len(data) / batch_size):
results.extend(predict.infer(list(rows), metadata))
print("Processed {} rows".format(len(results)))
end = datetime.datetime.now()
print("Time take to process {} rows is {}".format(len(data), end - start))
assert (len(results) == len(data))
print("Writing results to ", outputs)
with open(outputs, 'w') as f:
json.dump(results, f, indent=4)
print("First result is: ", results[0:1])
def generator():
path = "./image"
# 创建一个级联分类器
face_casecade = cv2.CascadeClassifier(
'haarcascade_frontalface_default.xml')
# 打开摄像头
camera = cv2.VideoCapture(0)
cv2.namedWindow('人脸识别')
while (True):
# 读取一帧图像
ret, frame = camera.read()
if ret:
# 转换为灰度图
gray_img = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# 人脸检测q
face = face_casecade.detectMultiScale(gray_img, 1.3, 5)
for (x, y, w, h) in face:
# 在原图上绘制矩形
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 0, 255), 2)
# 调整图像大小
new_frame = cv2.resize(gray_img[y:y + h, x:x + w], (64, 64))
# 识别l人脸
cv2.imwrite("test.png", new_frame)
res = infer("test.png")
cv2.putText(frame, res, (180, 320), cv2.FONT_HERSHEY_COMPLEX,
1, (180, 100, 255), 2, cv2.LINE_AA)
cv2.imshow('Dynamic', frame)
# 按下q键退出
if cv2.waitKey(100) & 0xff == ord('q'):
break
camera.release()
cv2.destroyAllWindows()
def post(self):
args = self.reqparse.parse_args()
data = args['data']
return predict.infer(list(data.values()), metadata)
def post(self):
args = self.reqparse.parse_args()
data = args['data']
return predict.infer(list(data.values()), types, tfidf_model,
svd_model)
def infer_from_model(model_name, messages):
_, params = build_model(params_path="./build/param",
enc_lstm_units=128)
model = tf.keras.models.load_model("./build/{}".format(model_name))
return infer(messages, model, params)