import argparse
import time

import cv2
import matplotlib.pyplot as plt
import numpy as np
from openvino.inference_engine import IECore


def parse_args() -> argparse.Namespace:
    """Parse and return command line arguments"""
    parser = argparse.ArgumentParser(add_help=False)
    args = parser.add_argument_group('Options')
    # fmt: off
    args.add_argument('-h', '--help', action='help', help='Show this help message and exit.')
    args.add_argument('-m', '--model', required=True, type=str,
                      help='Required. Path to an .xml or .onnx file with a trained model.')
    args.add_argument('-i', '--input', required=True, type=str, help='Required. Path to an image file.')
    args.add_argument('-d', '--device', default='CPU', type=str,
                      help='Optional. Specify the target device to infer on; CPU, GPU, MYRIAD, HDDL or HETERO: '
                           'is acceptable. The sample will look for a suitable plugin for device specified. '
                           'Default value is CPU.')
    # fmt: on
    return parser.parse_args()


def sample(model_location, image_location, device='CPU'):
    ie = IECore()

    net = ie.read_network(model=model_location)
    input_blob = next(iter(net.input_info))
    output_blob = next(iter(net.outputs))

    b, c, h, w = net.input_info[input_blob].input_data.shape
    image = cv2.imread(image_location)
    input_ratio = h / w
    target_ratio = image.shape[0] / image.shape[1]
    crop_axis = 0 if target_ratio > input_ratio else 1
    crop_factor = input_ratio * target_ratio / 2
    center = [image.shape[0] / 2, image.shape[1] / 2]
    x1 = int(center[0] - image.shape[0] * crop_factor) if crop_axis == 0 else 0
    x2 = int(center[0] + image.shape[0] * crop_factor) if crop_axis == 0 else image.shape[0]
    y1 = int(center[1] - image.shape[1] * crop_factor) if crop_axis == 1 else 0
    y2 = int(center[1] + image.shape[1] * crop_factor) if crop_axis == 1 else image.shape[1]
    # Crop to target aspect ratio
    image = image[x1:x2, y1:y2]
    if image.shape[:-1] != (h, w):
        image = cv2.resize(image, (w, h))

    image = image.transpose((2, 0, 1))
    # For batching
    image = np.expand_dims(image, axis=0)

    exec_net = ie.load_network(network=net, device_name=device)
    start = time.time()
    res = exec_net.infer(inputs={input_blob: image})
    print('First Inference Time Seconds: ' + str(time.time() - start))
    start = time.time()
    res = exec_net.infer(inputs={input_blob: image})
    print('Second Inference Time Seconds: ' + str(time.time() - start))
    start = time.time()
    res = exec_net.infer(inputs={input_blob: image})
    print('Third Inference Time Seconds: ' + str(time.time() - start))
    res = res[output_blob]
    depth = res[0][0]
    fig = plt.figure()
    ii = plt.imshow(depth, interpolation='nearest')
    fig.colorbar(ii)
    plt.show()


if __name__ == '__main__':
    parsed_args = parse_args()
    sample(parsed_args.model, parsed_args.input, parsed_args.device)