File size: 5,982 Bytes
95360c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
import numpy as np
import torch
from torch.utils.data import Dataset, DataLoader, Subset
from pathlib import Path
import os
import rasterio
import cv2
import pdb
from pyproj import Transformer


EXTENSIONS = (".jpg", ".jpeg", ".png", ".ppm", ".bmp", ".pgm", ".tif", ".tiff", ".webp")
ALL_BANDS = ['B01', 'B02', 'B03', 'B04', 'B05', 'B06', 'B07', 'B08', 'B09', 'B10', 'B11', 'B12', 'B8A']
S2A_BANDS = ['B01', 'B02', 'B03', 'B04', 'B05', 'B06', 'B07', 'B08', 'B09', 'B11', 'B12', 'B8A']
RGB_BANDS = ['B04', 'B03', 'B02']
S1_BANDS = ['VV', 'VH']

### SSL4EO stats
BAND_STATS = {
    'mean': {
        'B01': 1353.72696296,
        'B02': 1117.20222222,
        'B03': 1041.8842963,
        'B04': 946.554,
        'B05': 1199.18896296,
        'B06': 2003.00696296,
        'B07': 2374.00874074,
        'B08': 2301.22014815,
        'B8A': 2599.78311111,
        'B09': 732.18207407,
        'B10': 12.09952894,
        'B11': 1820.69659259,
        'B12': 1118.20259259,
        'VV': -12.54847273,
        'VH': -20.19237134
    },
    'std': {
        'B01': 897.27143653,
        'B02': 736.01759721,
        'B03': 684.77615743,
        'B04': 620.02902871,
        'B05': 791.86263829,
        'B06': 1341.28018273,
        'B07': 1595.39989386,
        'B08': 1545.52915718,
        'B8A': 1750.12066835,
        'B09': 475.11595216,
        'B10': 98.26600935,
        'B11': 1216.48651476,
        'B12': 736.6981037,
        'VV': 5.25697717,
        'VH': 5.91150917
    }
}

# BAND_STATS_S1 = {
#     'mean': {
#         'VV': -12.54847273,
#         'VH': -20.19237134
#     },
#     'std': {
#         'VV': 5.25697717,
#         'VH': 5.91150917
#     }
# }


def is_valid_file(filename):
    return filename.lower().endswith(EXTENSIONS)

def normalize(img, mean, std):
    min_value = mean - 2 * std
    max_value = mean + 2 * std
    img = (img - min_value) / (max_value - min_value) * 255.0
    img = np.clip(img, 0, 255).astype(np.uint8)
    #img = (img - min_value) / (max_value - min_value)
    #img = np.clip(img, 0, 1).astype(np.float32)
    return img

class EurosatDataset(Dataset):

    def __init__(self, root, bands='B2', split='train', transform=None, normalize=False, meta=False):
        self.root = Path(root,split)
        self.transform = transform
        if bands=='B13':
            self.bands = ALL_BANDS
        elif bands=='B12':
            self.bands = S2A_BANDS
        elif bands=='RGB':
            self.bands = RGB_BANDS
        elif bands=='B2':
            self.bands = S1_BANDS
            
        self.normalize = normalize
            
        self.classes = sorted([d.name for d in self.root.iterdir() if d.is_dir()])
        self.class_to_idx = {cls_name: i for i, cls_name in enumerate(self.classes)}

        self.samples = []
        self.targets = []

        #pdb.set_trace()
        for froot, _, fnames in sorted(os.walk(self.root, followlinks=True)):
            for fname in sorted(fnames):
                if is_valid_file(fname):
                    path = os.path.join(froot, fname)
                    self.samples.append(path)
                    target = self.class_to_idx[Path(path).parts[-2]]
                    self.targets.append(target)
        #print(self.root)
        #print(f"Found {len(self.samples)} images belonging to {len(self.classes)} classes")
        self.meta = meta

    def __getitem__(self, index):
        path = self.samples[index]
        target = self.targets[index]
        
        with rasterio.open(path) as f:
            if self.bands == ALL_BANDS:
                array = f.read().astype(np.int16)
            elif self.bands == S2A_BANDS:
                array = f.read((1,2,3,4,5,6,7,8,9,11,12,13)).astype(np.int16)
            elif self.bands == RGB_BANDS:
                array = f.read((4,3,2)).astype(np.int16)
            elif self.bands == S1_BANDS:
                array = f.read().astype(np.float32)
                            
            img = array.transpose(1, 2, 0)

            if self.meta:
                # get lon, lat, time
                cx,cy = f.xy(f.height // 2, f.width // 2)
                # convert to lon, lat
                crs_transformer = Transformer.from_crs(f.crs, 'epsg:4326')
                lon, lat = crs_transformer.transform(cx,cy)
                # no time
                meta_info = np.array([lon, lat, 0, 0]).astype(np.float32)
                #meta_info = np.array([0, 0, 0, 0]).astype(np.float32)
                #meta_info = np.array([np.nan, np.nan, np.nan, np.nan]).astype(np.float32)

        channels = []
        
        for i,b in enumerate(self.bands):
            ch = img[:,:,i]
            if self.normalize:
                ch = normalize(ch, mean=BAND_STATS['mean'][b], std=BAND_STATS['std'][b])
            elif self.bands == S2A_BANDS:
                ch = (ch / 10000.0 * 255.0).astype('uint8')

            if b=='B8A': # EuSAT band order is different than SSL4EO
                channels.insert(8,ch)
            else:
                channels.append(ch)
        #img = np.dstack(channels)
        img = np.stack(channels, axis=0).astype('float32') / 255.0

        if self.transform is not None:
            img = self.transform(img)

        if self.meta:
            return img, target, meta_info
        else:
            return img, target

    def __len__(self):
        return len(self.samples)


class Subset(Dataset):
    r"""
    Subset of a dataset at specified indices.

    Arguments:
        dataset (Dataset): The whole Dataset
        indices (sequence): Indices in the whole set selected for subset
    """
    def __init__(self, dataset, indices, transform=None):
        self.dataset = dataset
        self.indices = indices
        self.transform = transform

    def __getitem__(self, idx):
        im, target = self.dataset[self.indices[idx]]
        if self.transform:
            im = self.transform(im)
        return im, target

    def __len__(self):
        return len(self.indices)