如何恢复Tensorflow模型并预测输入图像

Question

我恢复了训练有素的张量流模型，然后我想要检测输入图像结果。 这里恢复代码：

import tensorflow as tf
import cv2 
sess=tf.Session() 
image_size=128   

saver = tf.train.import_meta_graph('my_test_model-1000.meta')
saver.restore(sess,tf.train.latest_checkpoint('./'))
sess.run(tf.global_variables_initializer())

然后如何使用此模型预测128 * 128 * 3 rgb输入图像？

Answer 1

import tensorflow as tf
import cv2 

def model(inputs):
    # You define your model over here
    ...
    ...
return logtis

image = cv2.imread("image_path.jpg")
ip_tensor = tf.placeholder(tf.float32, (None, 128, 128, 3)) 
logits = model(ip_tensor)
with tf.Session() as sess:
    saver.restore(sess, tf.train.latest_checkpoint('./'))  # restore your model
    feed={ip_tensor: inputs}  # prepare your feed to the saved model
    predictions = sess.run(tf.argmax(logits, 1), feed_dict=feed) # make prediction