esrgan_exp.py

import os
import time
from PIL import Image
import numpy as np
import tensorflow as tf
import tensorflow_hub as hub
import matplotlib.pyplot as plt
import h5py
from util.util import scale, descale, get_grid_values_all
from util.setup import home_dir

target_param = 'cld_opd_dcomp'

# SAVED_MODEL_PATH = "https://tfhub.dev/captain-pool/esrgan-tf2/1"
SAVED_MODEL_PATH = home_dir + '/esrgan-tf2_1'

model = hub.load(SAVED_MODEL_PATH)


def get_image_from_file(in_file):
    h5f = h5py.File(in_file, 'r')

    # s_x = slice(2622, 3134)
    # s_y = slice(2622, 3134)
    s_x = slice(2110, 3646)
    s_y = slice(2110, 3646)
    # s_x = slice(1854, 3902)
    # s_y = slice(1854, 3902)

    # bt = get_grid_values_all(h5f, 'temp_11_0um_nom')
    # bt = bt[s_y, s_x]

    cld_opd = get_grid_values_all(h5f, target_param)
    cld_opd = cld_opd[s_y, s_x]

    shape = cld_opd.shape
    data = cld_opd.flatten()

    lo = 0.0
    hi = 160.0

    data -= lo
    data /= (hi - lo)

    not_valid = np.isnan(data)
    data[not_valid] = 0

    data = np.reshape(data, shape)

    return data


def get_image(data):
    shape = data.shape
    data = data.flatten()

    lo = 0.0
    hi = 160.0

    data -= lo
    data /= (hi - lo)

    not_valid = np.isnan(data)
    data[not_valid] = 0

    data = np.reshape(data, shape)

    return data


def preprocess_image(hr_image):
  """ Loads image from path and preprocesses to make it model ready
      Args:
        image_path: Path to the image file
  """
  # hr_image = tf.image.decode_image(tf.io.read_file(image_path))
  # If PNG, remove the alpha channel. The model only supports
  # images with 3 color channels.
  if hr_image.shape[-1] == 4:
      hr_image = hr_image[..., :-1]
  elif len(hr_image.shape) == 2:
      tmp_image = np.zeros([hr_image.shape[0], hr_image.shape[1], 3])
      tmp_image[:, :, 0] = hr_image
      tmp_image[:, :, 1] = hr_image
      tmp_image[:, :, 2] = hr_image
      hr_image = tmp_image
      hr_image *= 255.0

  hr_size = (tf.convert_to_tensor(hr_image.shape[:-1]) // 4) * 4
  hr_image = tf.image.crop_to_bounding_box(hr_image, 0, 0, hr_size[0], hr_size[1])
  hr_image = tf.cast(hr_image, tf.float32)

  return tf.expand_dims(hr_image, 0)


def run(in_file, out_file):
    # model = hub.load(SAVED_MODEL_PATH)
    t0 = time.time()
    hr_image = get_image_from_file(in_file)
    # hr_image = get_image(in_file)
    hr_image = preprocess_image(hr_image)
    t1 = time.time()
    print('processing time: ', (t1-t0))
    print('start inference: ', hr_image.shape)
    t0 = time.time()
    fake_image = model(hr_image)
    t1 = time.time()
    print('inference time: ', (t1-t0))
    fake_image = tf.squeeze(fake_image)

    if out_file is not None:
        np.save(out_file, fake_image)
    else:
        return fake_image, hr_image