模型预测报错：is not an element of this graph（已解决）

……
model = ResNet50(weights='imagenet')
……
def main():
    x = model.predict(x)

……
import tensorflow as tf
from tensorflow.python.keras.backend import set_session
from tensorflow.python.keras.models import load_model
tf_config = tf.ConfigProto()
sess = tf.Session(config=tf_config)
graph = tf.get_default_graph()

set_session(sess)
model = ResNet50(weights='imagenet')
……
def main():
    global sess
    global graph
    with graph.as_default():
        set_session(sess)
        x = model.predict(x)