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 

28from torch.utils.data import Dataset 

29import numpy as np 

30import torch 

31 

32 

33class S1SLC(Dataset): 

34 r""" 

35 The Polarimetric SAR dataset with the labels provided by 

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

37 

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

39 

40 Arguments: 

41 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 

42 transform : the transform applied the cropped image 

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

44 

45 Note: 

46 An example usage : 

47 

48 .. code-block:: python 

49 

50 import torchcvnn 

51 from torchcvnn.datasets import S1SLC 

52 

53 def transform(patches): 

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

55 # S1SLC provides the dual HH, HV polarizations 

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

57 return np.stack(patches) 

58 

59 dataset = S1SLC(rootdir, transform=transform 

60 X, y = dataset[0] 

61 

62 """ 

63 

64 def __init__(self, root, transform=None, lazy_loading=True): 

65 self.transform = transform 

66 self.lazy_loading = lazy_loading 

67 # Get list of subfolders in the root path 

68 subfolders = [ 

69 os.path.join(root, name) 

70 for name in os.listdir(root) 

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

72 ] 

73 

74 self.data = [] 

75 self.labels = [] 

76 

77 for subfolder in subfolders: 

78 # Define paths to the .npy files 

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

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

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

82 

83 # Load the .npy files 

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

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

86 

87 if not lazy_loading: 

88 # If not lazy loading, we load all the data in main memory 

89 # Concatenate HH and HV to create a two-channel array 

90 data = np.stack((hh, hv), axis=1) # Shape: (B, 2, H, W) 

91 else: 

92 # If lazy loading, we store the paths to the .npy files 

93 num_patches = hh.shape[0] 

94 data = [ 

95 (hh_path, hv_path, patch_idx) for patch_idx in range(num_patches) 

96 ] 

97 

98 # For the labels, we can preload everything in main memory 

99 label = np.load(labels_path, mmap_mode="r") 

100 label = [int(l.item()) - 1 for l in label] # Convert to 0-indexed labels 

101 

102 # Append data and labels to the lists 

103 self.data.extend(data) 

104 self.labels.extend(label) 

105 

106 self.classes = list(set(self.labels)) 

107 

108 def __len__(self): 

109 return len(self.data) 

110 

111 def __getitem__(self, idx): 

112 

113 if self.lazy_loading: 

114 hh_path, hv_path, patch_idx = self.data[idx] 

115 

116 # Load the .npy files 

117 hh = np.load(hh_path) 

118 hv = np.load(hv_path) 

119 

120 # Extract the right patch 

121 hh_patch = hh[patch_idx] 

122 hv_patch = hv[patch_idx] 

123 

124 # Concatenate HH and HV to create a two-channel array 

125 image = np.stack((hh_patch, hv_patch), axis=0) # Shape: (2, H, W) 

126 else: 

127 image = self.data[idx] 

128 

129 label = self.labels[idx] 

130 

131 if self.transform: 

132 image = self.transform(image) 

133 

134 return image, label