def scao_model_iv(AI_saved_dir, arti_flux_678, arti_label_678):
    #-----------------------------------
    # Load AI
    print('Loading AI...')
    width_of_data = 1
    img_maj = 3
    image_shape = (width_of_data, img_maj)
    kernal_shape = (width_of_data, 2)
    num_kernal_1 = 32
    num_kernal_2 = 64
    num_conn_neural = 100
    num_label = 3
    #-----------------------------------
    # Construct an AI
    tf.reset_default_graph()
    x = tf.placeholder(tf.float32, [None, width_of_data * img_maj], name='x')
    y_true = tf.placeholder(tf.float32, [None, 3], name='y_true')
    y_true_cls = tf.argmax(y_true, axis=1)
    x_image = tf.reshape(x, [-1, image_shape[0], image_shape[1], 1])
    # First layer( First kernal)
    W_conv1 = weight_variable(
        [kernal_shape[0], kernal_shape[1], 1, num_kernal_1])
    b_conv1 = bias_variable([num_kernal_1])
    h_conv1 = tf.nn.selu(
        tf.nn.conv2d(x_image, W_conv1, [1, 1, 1, 1], 'SAME') + b_conv1)
    # Second layer( Second kernal)
    W_conv2 = weight_variable(
        [kernal_shape[0], kernal_shape[1], num_kernal_1, num_kernal_2])
    b_conv2 = bias_variable([num_kernal_2])
    h_conv2 = tf.nn.selu(
        tf.nn.conv2d(h_conv1, W_conv2, [1, 1, 1, 1], 'SAME') + b_conv2)
    # Third layer ( Fully connected)
    W_fc1 = weight_variable(
        [image_shape[0] * image_shape[1] * num_kernal_2, num_conn_neural])
    b_fc1 = bias_variable([num_conn_neural])
    h_conv2_flat = tf.reshape(
        h_conv2, [-1, image_shape[0] * image_shape[1] * num_kernal_2])
    h_fc1 = tf.nn.selu(tf.matmul(h_conv2_flat, W_fc1) + b_fc1)
    # Output layer
    W_fc2 = weight_variable([num_conn_neural, num_label])
    b_fc2 = bias_variable([num_label])
    layer_last = tf.matmul(h_fc1, W_fc2) + b_fc2
    y_pred = tf.nn.softmax(layer_last)
    y_pred_cls = tf.argmax(y_pred, axis=1)
    correct_prediction = tf.equal(y_pred_cls, y_true_cls)
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    # Saver
    saver = tf.train.Saver()
    print("AI:{0}".format(AI_saved_dir))
    if not os.path.exists(AI_saved_dir):
        print("No AI can be restore, please check folder ./checkpoints")
        exit(1)
    save_path = os.path.join(AI_saved_dir, 'best_validation')
    session = tf.Session()
    # Restore previous weight
    saver.restore(sess=session, save_path=save_path)

    #-----------------------------------
    # Make a prediction
    def predict_label(images, labels):
        # Number of images.
        num_images = len(images)
        # initialize
        label_pred = np.zeros(num_images * 3).reshape((num_images, 3))
        feed_dict = {x: images[:], y_true: labels[:]}
        # process
        label_pred = session.run(y_pred, feed_dict=feed_dict)
        return label_pred

    label_pred_678 = predict_label(arti_flux_678, arti_label_678)
    #-----------------------------------
    # Close session
    session.close()
    return label_pred_678
 img_maj = 3
 image_shape = (width_of_data, img_maj)
 kernal_shape = (width_of_data, 2)
 num_kernal_1 = 32
 num_kernal_2 = 64
 num_conn_neural = 100
 num_label = 3
 #-----------------------------------
 # Construct an AI
 tf.reset_default_graph()
 x = tf.placeholder(tf.float32, [None, width_of_data * img_maj], name='x')
 y_true = tf.placeholder(tf.float32, [None, 3], name='y_true')
 y_true_cls = tf.argmax(y_true, axis=1)
 x_image = tf.reshape(x, [-1, image_shape[0], image_shape[1], 1])
 # First layer( First kernal)
 W_conv1 = weight_variable(
     [kernal_shape[0], kernal_shape[1], 1, num_kernal_1])
 b_conv1 = bias_variable([num_kernal_1])
 h_conv1 = tf.nn.selu(
     tf.nn.conv2d(x_image, W_conv1, [1, 1, 1, 1], 'SAME') + b_conv1)
 # Second layer( Second kernal)
 W_conv2 = weight_variable(
     [kernal_shape[0], kernal_shape[1], num_kernal_1, num_kernal_2])
 b_conv2 = bias_variable([num_kernal_2])
 h_conv2 = tf.nn.selu(
     tf.nn.conv2d(h_conv1, W_conv2, [1, 1, 1, 1], 'SAME') + b_conv2)
 # Third layer ( Fully connected)
 W_fc1 = weight_variable(
     [image_shape[0] * image_shape[1] * num_kernal_2, num_conn_neural])
 b_fc1 = bias_variable([num_conn_neural])
 h_conv2_flat = tf.reshape(
     h_conv2, [-1, image_shape[0] * image_shape[1] * num_kernal_2])