DBPN-Tensorflow/DataPreprocessing.py at main · TzeLun/DBPN-Tensorflow · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
from Data_Eval import crop, Bicubic_LR_to_HR

# Dataset used -> DIV2K: https://ieeexplore.ieee.org/document/8014884
# Dataset obtained from https://data.vision.ee.ethz.ch/cvl/DIV2K/

# Folders containing original images from the DIV2K
list_to_convert = ["C:/Users/Tze Lun/DIV2K/x_train/X4/",
                   "C:/Users/Tze Lun/DIV2K/x_valid/X4/",
                   "C:/Users/Tze Lun/DIV2K/y_train/HR/",
                   "C:/Users/Tze Lun/DIV2K/y_valid/HR/"]

# Folders to contain the cropped images
list_to_save = ["C:/Users/Tze Lun/DIV2K/x_train/X4_40x40/",
                "C:/Users/Tze Lun/DIV2K/x_valid/X4_40x40/",
                "C:/Users/Tze Lun/DIV2K/y_train/HR_160x160/",
                "C:/Users/Tze Lun/DIV2K/y_valid/HR_160X160/"]

# Folders containing 5 test image from Set5, available on author's Github:
# https://github.com/alterzero/DBPN-Pytorch/tree/master/Input/Set5_LR_x4
dir_test_file = "C:/Users/Tze Lun/DIV2K/x_test/LR_original/"
dir_save = "C:/Users/Tze Lun/DIV2K/x_test/X4_40X40/"

# Converting LR to HR with bicubic interpolation
LR_test_data_path = "C:/Users/Tze Lun/DIV2K/x_test/ModelEval/"
LR_test_data_save_path = "C:/Users/Tze Lun/DIV2K/x_test/ScaledLR/"
# re-crop the images to the desired size
for i in range(len(list_to_convert)):
    if i <= 1:
        size = (40, 40)  # width, height
        crop(list_to_convert[i], list_to_save[i], size)
    else:
        size = (160, 160)
        crop(list_to_convert[i], list_to_save[i], size)

size = (40, 40)
crop(dir_test_file, dir_save, size)

# Upscaling the LR test data to obtain HR counterparts
Bicubic_LR_to_HR(LR_test_data_path, LR_test_data_save_path, 4)