print("TRAIN IMAGE #"+str(train_batch_size*batchs))
if epoch_over is True:
print("END EPOCH "+str(epochs))
end = timer()
total_batchs += batchs
print("TRAIN TIME ELAPSED "+str(end-start))
###test on unseen data, init variables
print("BEGIN TESTING")
start = timer()
batchs = 0
epoch_over = False
accuracy = []
###begin testing/validation loop, use entire validation dataset
while not epoch_over:
test_batch, one_hots, epoch_over = trainer.get_test_batch(test_batch_size)
#get predicted class of test data from model
prediction = sess.run([net.output], feed_dict={net.inputs:test_batch})[0]
accuracy.append(np.mean(np.equal(np.argmax(prediction, axis=1), np.argmax(one_hots, axis=1)).astype("float32")))
batchs += 1
if batchs % 5 == 0:
print("TEST IMAGE #"+str(test_batch_size*batchs))
if epoch_over is True:
print("TEST EPOCH OVER")
epoch_over = False
batchs = 0
test_accuracy = np.mean(accuracy)
print("TEST ACCURACY "+str(test_accuracy))
write_to_tensorboard("Accuracy/Test Data", test_accuracy, epochs, summary_writer)
end = timer()
print("TEST TIME ELAPSED "+str(end-start))
