accuracy = 0
with tf.Session() as sess:
sess.run(init)
actual_steps = 0
runs = 0
index = 0
testlength = int(len(itest) / 70)
saver.restore(sess, var_export_dir)
for i in range((actual_steps - runs), steps):
matches = tf.equal(tf.argmax(y_pred, 1), tf.argmax(y_true, 1))
test_x, test_y, index = nextImageBatch(itest, testlength,
len(subfolders), index)
acc = tf.reduce_mean(tf.cast(matches, tf.float32))
print("ACCURACY: ")
get_acc = sess.run(acc,
feed_dict={
x_image: test_x,
y_true: test_y,
hold_prob: 1.0
})
print(get_acc)
test_saver.write(str(get_acc) + '\n')
accuracy = accuracy + get_acc
#saver.restore(sess, var_export_dir)
for i in range((actual_steps-runs), steps):
# We will check time required to run each step using start_time and end_time
train_time_saver = open('<PATH>/TempFM2/train_time.txt', 'a+')
start_time = time.time()
print('Saving step number');
step_saver.write(str(i) + '\n')
step_saver.close()
step_saver = open('<PATH>/TempFM2/steps.txt', 'a+')
batch_x, batch_y, index = nextImageBatch(itrain, trainlength, len(subfolders), index)
sess.run(train, feed_dict={x_image:batch_x, y_true:batch_y, pool_prob:0.75, hold_prob:0.5})
#print(trainlength)
#print accuracy every few steps
if (i+1)%10==0:
i1 = 0
print("ON STEP {}".format(i))
matches = tf.equal(tf.argmax(y_pred, 1), tf.argmax(y_true, 1))
test_x, test_y, i1 = nextImageBatch(itest, testlength, len(subfolders), i1)
acc = tf.reduce_mean(tf.cast(matches, tf.float32))
print("ACCURACY: ")
accuracy = sess.run(acc, feed_dict={x_image:test_x, y_true:test_y, pool_prob:1.0, hold_prob:1.0})
print(accuracy)