File size: 6,506 Bytes
a16f583
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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

import numpy as np
import torch
from torch.utils.data import Dataset, DataLoader
import os
from pathlib import Path
import re
import glob


class RasterTensorDatasetMapping(Dataset):
    def __init__(self, base_path):
        """
        Initialize the dataset

        Parameters:
        base_path: str, base path to RasterTensorData directory
        subfolder: str, name of the subfolder (e.g., 'Elevation')
        """
        # Replace "OC_LUCAS_LFU_LfL_Coordinates" with "RasterTensorData" in the base path
        self.base_path = Path(base_path.replace("Coordinates1Mil", "RasterTensorData"))
        self.folder_path = self.base_path

        # Create ID to filename mapping
        self.id_to_file = self._create_id_mapping()

        # Load all numpy arrays into memory (optional, can be modified to load on demand)
        self.data_cache = {}
        for id_num, filepath in self.id_to_file.items():
            self.data_cache[id_num] = np.load(filepath)

    def _create_id_mapping(self):
        """Create a dictionary mapping IDs to their corresponding file paths"""
        id_to_file = {}

        for file_path in self.folder_path.glob("*.npy"):
            # Extract ID number from filename
            match = re.search(r'ID(\d+)N', file_path.name)
            if match:
                id_num = int(match.group(1))
                id_to_file[id_num] = file_path

        return id_to_file

    def get_tensor_by_location(self, id_num, x, y, window_size=17):
        """
        Get a window_size x window_size square around the specified x,y coordinates

        Parameters:
        id_num: int, ID number from filename
        x: int, x coordinate
        y: int, y coordinate
        window_size: int, size of the square window (default 17)

        Returns:
        torch.Tensor: window_size x window_size tensor
        """
        if id_num not in self.id_to_file:
            raise ValueError(f"ID {id_num} not found in dataset")

        # Get the data array
        if id_num in self.data_cache:
            data = self.data_cache[id_num]
        else:
            data = np.load(self.id_to_file[id_num])

        # Calculate window boundaries
        half_window = window_size // 2
        x_start = int(max(0, x - half_window))
        x_end = int(min(data.shape[0], x + half_window + 1))
        y_start = int(max(0, y - half_window))
        y_end = int(min(data.shape[1], y + half_window + 1))

        # Extract window
        window = data[x_start:x_end, y_start:y_end]

        # Pad if necessary
        if window.shape != (window_size, window_size):
            padded_window = np.zeros((window_size, window_size))
            x_offset = half_window - (x - x_start)
            y_offset = half_window - (y - y_start)
            padded_window[
                x_offset:x_offset+window.shape[0],
                y_offset:y_offset+window.shape[1]
            ] = window
            window = padded_window

        return torch.from_numpy(window).float()

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

    def __getitem__(self, idx):
        # This is a placeholder implementation
        # Modify according to your specific needs
        id_num = list(self.id_to_file.keys())[idx]
        return self.data_cache[id_num]

# Example usage:
"""
# Initialize the dataset
base_path = "/content/drive/MyDrive/Colab Notebooks/MappingSOC/Data/RasterTensorData"
dataset = RasterTensorDataset(base_path, "Elevation")

# Get the dictionary mapping IDs to filenames
id_mapping = dataset.id_to_file
print("ID to filename mapping:", id_mapping)

# Get a 17x17 window for a specific location
id_num = 10  # example ID
x, y = 100, 100  # example coordinates
window = dataset.get_tensor_by_location(id_num, x, y)
print("Window shape:", window.shape)

# Create a DataLoader if needed
from torch.utils.data import DataLoader
dataloader = DataLoader(dataset, batch_size=4, shuffle=True)
"""


class MultiRasterDatasetMapping(Dataset):
    def __init__(self, subfolders, dataframe):
        """
        Parameters:
        subfolders: list of str, names of subfolders to include
        dataframe: pandas.DataFrame, contains columns GPS_LONG, GPS_LAT, and OC (target variable)
        """
        self.subfolders = subfolders
        self.dataframe = dataframe
        self.datasets = {
            subfolder: RasterTensorDatasetMapping(subfolder)
            for subfolder in subfolders
        }
        self.coordinates = {
            subfolder: np.load(f"{subfolder}/coordinates.npy")
            for subfolder in subfolders
        }

    def find_coordinates_index(self, subfolder, longitude, latitude):
        """
        Finds the index of the matching coordinates in the subfolder's coordinates.npy file.

        Parameters:
        subfolder: str, name of the subfolder
        longitude: float, longitude to match
        latitude: float, latitude to match

        Returns:
        tuple: (id_num, x, y) if match is found, otherwise raises an error
        """
        coords = self.coordinates[subfolder]
        # Assuming the first two columns of `coordinates.npy` are longitude and latitude
        match = np.where((coords[:, 1] == longitude) & (coords[:, 0] == latitude))[0]
        if match.size == 0:
            raise ValueError(f"Coordinates ({longitude}, {latitude}) not found in {subfolder}")

        # Return id_num, x, y from the same row
        return coords[match[0], 2], coords[match[0], 3], coords[match[0], 4]

    def __getitem__(self, index):
        """
        Retrieve tensor and target value for a given index.

        Parameters:
        index: int, index of the row in the dataframe

        Returns:
        tuple: (tensor, OC), where tensor is the data and OC is the target variable
        """
        row = self.dataframe.iloc[index]
        longitude, latitude = row["longitude"], row["latitude"]

        tensors = {}
        for subfolder in self.subfolders:
            id_num, x, y = self.find_coordinates_index(subfolder, longitude, latitude)
            tensors[subfolder] = self.datasets[subfolder].get_tensor_by_location(id_num, x, y)

        return longitude, latitude, tensors

    def __len__(self):
        """
        Return the number of samples in the dataset.
        """
        return len(self.dataframe)

    def get_tensor_by_location(self, subfolder, id_num, x, y):
        """Get tensor from specific subfolder dataset"""
        return self.datasets[subfolder].get_tensor_by_location(id_num, x, y)