create initial structure

mu_map/data/__init__.py

+from .datasets import *

mu_map/data/datasets.py

+import os
+
+import pandas as pd
+import pydicom
+from torch.utils.data import Dataset
+
+class MuMapDataset(Dataset):
+
+    def __init__(self):
+        super().__init__()
+
+        # read csv file and from that access dicom files
+
+    def __getitem__(self, index):
+        pass
+
+    def __len__(self):
+        pass
+
+
+__all__ = [MuMapDataset.__name__]

mu_map/data/preprocessing.py

+import torch
+
+
+def norm_max(tensor: torch.Tensor):
+    return (tensor - tensor.min()) / (tensor.max() - tensor.min())
+
+
+class MaxNorm:
+    def __call__(self, tensor: torch.Tensor):
+        return norm_max(tensor)
+
+
+def norm_mean(tensor: torch.Tensor):
+    return tensor / tensor.mean()
+
+
+class MeanNorm:
+    def __call__(self, tensor: torch.Tensor):
+        return norm_mean(tensor)
+
+
+def norm_gaussian(tensor: torch.Tensor):
+    return (tensor - tensor.mean()) / tensor.std()
+
+
+class GaussianNorm:
+    def __call__(self, tensor: torch.Tensor):
+        return norm_gaussian(tensor)
+
+
+__all__ = [
+    norm_max.__name__,
+    norm_mean.__name__,
+    norm_gaussian.__name__,
+    MaxNorm.__name__,
+    MeanNorm.__name__,
+    GaussianNorm.__name__,
+]

mu_map/test.py

+import torch
+
+from .data.preprocessing import *
+
+means = torch.full((10, 10, 10), 5.0)
+stds = torch.full((10, 10, 10), 10.0)
+x = torch.normal(means, stds)
+
+print(f"Before: mean={x.mean():.3f} std={x.std():.3f}")
+
+y = norm_gaussian(x)
+print(f" After: mean={y.mean():.3f} std={y.std():.3f}")
+y = GaussianNorm()(x)
+print(f" After: mean={y.mean():.3f} std={y.std():.3f}")
+
+
+
+