eval.py

'''
A Convolutional Network implementation example using TensorFlow library.
This example is using the MNIST database of handwritten digits
(http://yann.lecun.com/exdb/mnist/)

Author: Aymeric Damien
Project: https://github.com/aymericdamien/TensorFlow-Examples/
'''

from __future__ import print_function

import tensorflow as tf

# Import MNIST data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/mnist", one_hot=True)

# Launch the graph
with tf.Session() as sess:
    tf.saved_model.loader.load(sess, ["EVALUATING"], "/model/cnn_model")
    # Retrieve placeholder tensor by name
    graph = tf.get_default_graph()
    x = graph.get_tensor_by_name("x:0")
    y = graph.get_tensor_by_name("y:0")
    accuracy = graph.get_tensor_by_name("accuracy:0")
    keep_prob = graph.get_tensor_by_name("keep_prob:0")

    # Calculate accuracy for 256 mnist test images
    print("Testing Accuracy:", \
        sess.run(accuracy, feed_dict={x: mnist.test.images[:256],
                                      y: mnist.test.labels[:256],
                                      keep_prob: 1.}))