pixel-smasher/recreate_cropped.py at master · ekcomputer/pixel-smasher · 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
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
import numpy as np
from multiprocessing import Pool
import multiprocessing
import rasterio as rio
import glob as gl
import cv2
import os
import sys

cropped_folder = '/data_dir/classified_shield_v2/008_ESRGAN_x10_PLANET_noPreTrain_130k_Test_hold_shield_v2/visualization/HR/x10/'
cropped_suffix = 'T0' #'008_ESRGAN_x10_PLANET_noPreTrain_130k_Test'

uncropped_folder = '/data_dir/Scenes-shield-gt/'

save_folder = '/data_dir/classified_shield_gt_georef/HR/10x/'

img_list = gl.glob(uncropped_folder + '*SR.tif')

def main():
    """A multi-thread tool to put cropped images back into full (non-cropped) array based on index."""

    #n_thread = multiprocessing.cpu_count() #1

    crop_sz = 480 # num px in x and y
    step = 240
    thres_sz = 48

    #pool = Pool(4) # (n_thread)
    #for path in img_list:
    #    pool.apply_async(worker,
                         #args=(path, SR_folder, cropped_suffix, save_folder, crop_sz, step, thres_sz))
    #pool.close()
    #pool.join()
    #print('All subprocesses done.')

    for path in img_list:
        worker(path, cropped_folder, cropped_suffix, save_folder, crop_sz, step, thres_sz)

def worker(uncropped_path, cropped_folder, cropped_suffix, save_folder, crop_sz, step, thres_sz):
    '''
    This worker works on one uncropped image at a time to get the correct indexes for each cropped subset of the uncropped image,
    grab that cropped subset, and put it into a new array.

    '''
    # load uncropped image:
    uncropped_name = os.path.basename(uncropped_path)
    uncropped_img = cv2.imread(uncropped_path, cv2.IMREAD_UNCHANGED)

    uncropped_img_rio = rio.open(uncropped_path)
    profile = uncropped_img_rio.profile
    profile.update(nodata = 255)

    #get output size:
    cropped_path = gl.glob(cropped_folder + '*.png')[0]
    cropped_ex = cv2.imread(cropped_path, cv2.IMREAD_UNCHANGED)

    dtype = str(cropped_ex.dtype)
    if dtype == 'uint8':
        dtype = rio.uint8
    if dtype == 'float64':
        dtype = rio.float64
    if dtype == 'bool':
        dtype = rio.bool
    if dtype == 'uint16':
        dtype = rio.uint16
    if dtype == 'int16':
        dtype = rio.int16
    if dtype == 'uint32':
        dtype = rio.uint32
    if dtype == 'int32':
        dtype = rio.int32
    if dtype == 'float32':
        dtype = rio.float32

    n_channels = len(cropped_ex.shape)
    if n_channels == 2:
        h, w = uncropped_img.shape[0:2]
        new_uncropped_img = np.zeros((h,w,1))
        new_uncropped_img.fill(255)
        profile.update(dtype = dtype, count = 1)
    elif n_channels == 3:
        h, w, c = uncropped_img.shape
        new_uncropped_img = np.zeros((h,w,3))
        new_uncropped_img.fill(255)
        profile.update(dtype = dtype, count = 3)
    else:
        raise ValueError('Wrong image shape - {}'.format(n_channels))

    h_space = np.arange(0, h - crop_sz + 1, step)
    if h - (h_space[-1] + crop_sz) > thres_sz:
        h_space = np.append(h_space, h - crop_sz)
    w_space = np.arange(0, w - crop_sz + 1, step)
    if w - (w_space[-1] + crop_sz) > thres_sz:
        w_space = np.append(w_space, w - crop_sz)

    index = 1
    for x in h_space:
        for y in w_space:
            if n_channels == 2:
                crop_img = uncropped_img[x:x + crop_sz, y:y + crop_sz]
            else:
                crop_img = uncropped_img[x:x + crop_sz, y:y + crop_sz, :]
            crop_img = np.ascontiguousarray(crop_img)
            if ~np.any(np.sum(crop_img,axis=2)==0): # if all three bands == 0

                cropped_path = str(cropped_folder) + uncropped_name.replace('.tif', '_s{:04d}_'.format(index)) + str(cropped_suffix) + '.png'
                #print(cropped_path)
                #print(os.path.isfile(cropped_path))

                try:
                    cropped_img = cv2.imread(cropped_path, cv2.IMREAD_UNCHANGED)
                    if len(np.unique(cropped_img)) == 2:
                        cropped_img[cropped_img < 1] = 0
                        cropped_img[cropped_img > 1] = 1
                except:
                    cropped_img = np.zeros(crop_img.shape)
                    cropped_img.fill(255)

                if n_channels == 2:
                    new_uncropped_img[x:x + crop_sz, y:y + crop_sz, 0] = cropped_img[:]
                else:
                    new_uncropped_img[x:x + crop_sz, y:y + crop_sz, :] = cropped_img[:]

                index += 1
            else:
                #print('all zero')
                pass

    save_path = save_folder + uncropped_name

    with rio.Env():
        with rio.open(save_path, 'w', **profile) as dst:
            new_uncropped_img = np.rollaxis(new_uncropped_img, 2)
            dst.write(new_uncropped_img.astype(dtype))

    return 'Processing {:s} ...'.format(uncropped_name)


if __name__ == '__main__':
    main()