Coverage for  / home / runner / work / torchcvnn / torchcvnn / src / torchcvnn / datasets / s1slc.py: 0%

1# MIT License 

2 

3# Copyright (c) 2025 Quentin Gabot 

4 

5# Permission is hereby granted, free of charge, to any person obtaining a copy 

6# of this software and associated documentation files (the "Software"), to deal 

7# in the Software without restriction, including without limitation the rights 

8# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 

9# copies of the Software, and to permit persons to whom the Software is 

10# furnished to do so, subject to the following conditions: 

11 

12# The above copyright notice and this permission notice shall be included in 

13# all copies or substantial portions of the Software. 

14 

15# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 

16# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 

17# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 

18# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 

19# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 

20# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 

21# SOFTWARE. 

22 

23 

24# Standard imports 

25import os 

26 

27# External imports 

28import numpy as np 

29from torch.utils.data import Dataset 

30 

31 

32class S1SLC(Dataset): 

33 r""" 

34 The Polarimetric SAR dataset with the labels provided by 

35 https://ieee-dataport.org/open-access/s1slccvdl-complex-valued-annotated-single-look-complex-sentinel-1-sar-dataset-complex 

36 

37 We expect the data to be already downloaded and available on your drive. 

38 

39 Arguments: 

40 root: the top root dir where the data are expected. The data should be organized as follows: Sao Paulo/HH.npy, Sao Paulo/HV.npy, Sao Paulo/Labels.npy, Houston/HH.npy, Houston/HV.npy, Houston/Labels.npy, Chicago/HH.npy, Chicago/HV.npy, Chicago/Labels.npy 

41 transform : the transform applied the cropped image 

42 lazy_loading : if True, the data is loaded only when requested. If False, the data is loaded at the initialization of the dataset. 

43 

44 Note: 

45 An example usage : 

46 

47 .. code-block:: python 

48 

49 import torchcvnn 

50 from torchcvnn.datasets import S1SLC 

51 

52 def transform(patches): 

53 # If you wish, you could filter out some polarizations 

54 # S1SLC provides the dual HH, HV polarizations 

55 patches = [np.abs(patchi) for _, patchi in patches.items()] 

56 return np.stack(patches) 

57 

58 dataset = S1SLC(rootdir, transform=transform 

59 X, y = dataset[0] 

60 

61 """ 

62 

63 def __init__(self, root, transform=None, lazy_loading=False): 

64 if lazy_loading is True: 

65 raise DeprecationWarning( 

66 "Lazy loading is no longer supported for S1SLC dataset." 

67 ) 

68 

69 self.transform = transform 

70 # Get list of subfolders in the root path 

71 subfolders = [ 

72 os.path.join(root, name) 

73 for name in os.listdir(root) 

74 if os.path.isdir(os.path.join(root, name)) 

75 ] 

76 

77 self.data = {} 

78 self.classes = set() 

79 

80 for subfolder in subfolders: 

81 # Define paths to the .npy files 

82 hh_path = os.path.join(subfolder, "HH.npy") 

83 hv_path = os.path.join(subfolder, "HV.npy") 

84 labels_path = os.path.join(subfolder, "Labels.npy") 

85 

86 # Load the .npy files (using a memory map for memory efficiency) 

87 hh = np.load(hh_path, mmap_mode="r") 

88 hv = np.load(hv_path, mmap_mode="r") 

89 

90 # Load labels and convert to 0-indexed 

91 labels = np.load(labels_path, mmap_mode="r") 

92 labels = labels.astype(int).squeeze() - 1 

93 

94 # put labels in the set of classes 

95 self.classes.update(list(labels)) 

96 

97 self.data[subfolder] = { 

98 "hh": hh, 

99 "hv": hv, 

100 "labels": labels, 

101 } 

102 

103 self.classes = [int(c) for c in self.classes] 

104 

105 def __len__(self): 

106 return sum(len(city_data["labels"]) for city_data in self.data.values()) 

107 

108 def find_city_and_local_idx(self, idx): 

109 """ 

110 Given a global index, find the corresponding city and local index within that city's data. 

111 This is done in O(n_cities) time, which is acceptable since n_cities=3 

112 """ 

113 cumulative = 0 

114 for city, city_data in self.data.items(): 

115 city_size = len(city_data["labels"]) 

116 if idx < cumulative + city_size: 

117 local_idx = idx - cumulative 

118 return city, local_idx 

119 cumulative += city_size 

120 raise IndexError("Index out of range") 

121 

122 def __getitem__(self, idx): 

123 city, local_idx = self.find_city_and_local_idx(idx) 

124 

125 image = np.stack( 

126 [ 

127 self.data[city]["hh"][local_idx], 

128 self.data[city]["hv"][local_idx], 

129 ] 

130 ) 

131 label = self.data[city]["labels"][local_idx] 

132 

133 if self.transform: 

134 image = self.transform(image) 

135 

136 return image, label