--- a
+++ b/sybil/loaders/image_loaders.py
@@ -0,0 +1,76 @@
+from sybil.loaders.abstract_loader import abstract_loader
+import cv2
+import torch
+import pydicom
+from pydicom.pixel_data_handlers.util import apply_modality_lut
+import numpy as np
+
+LOADING_ERROR = "LOADING ERROR! {}"
+
+
+class OpenCVLoader(abstract_loader):
+
+    def load_input(self, path):
+        """
+        loads as grayscale image
+        """
+        return {"input": cv2.imread(path, 0)}
+
+    @property
+    def cached_extension(self):
+        return ".png"
+
+
+class DicomLoader(abstract_loader):
+    def __init__(self, cache_path, augmentations, args, apply_augmentations=True):
+        super(DicomLoader, self).__init__(cache_path, augmentations, args, apply_augmentations)
+        self.window_center = -600
+        self.window_width = 1500
+
+    def load_input(self, path):
+        try:
+            dcm = pydicom.dcmread(path)
+            dcm = apply_modality_lut(dcm.pixel_array, dcm)
+            arr = apply_windowing(dcm, self.window_center, self.window_width)
+            arr = arr//256  # parity with images loaded as 8 bit
+        except Exception:
+            raise Exception(LOADING_ERROR.format("COULD NOT LOAD DICOM."))
+        return {"input": arr}
+
+    @property
+    def cached_extension(self):
+        return ""
+
+
+def apply_windowing(image, center, width, bit_size=16):
+    """Windowing function to transform image pixels for presentation.
+    Must be run after a DICOM modality LUT is applied to the image.
+    Windowing algorithm defined in DICOM standard:
+    http://dicom.nema.org/medical/dicom/2020b/output/chtml/part03/sect_C.11.2.html#sect_C.11.2.1.2
+    Reference implementation:
+    https://github.com/pydicom/pydicom/blob/da556e33b/pydicom/pixel_data_handlers/util.py#L460
+    Args:
+        image (ndarray): Numpy image array
+        center (float): Window center (or level)
+        width (float): Window width
+        bit_size (int): Max bit size of pixel
+    Returns:
+        ndarray: Numpy array of transformed images
+    """
+    y_min = 0
+    y_max = 2 ** bit_size - 1
+    y_range = y_max - y_min
+
+    c = center - 0.5
+    w = width - 1
+
+    below = image <= (c - w / 2)  # pixels to be set as black
+    above = image > (c + w / 2)  # pixels to be set as white
+    between = np.logical_and(~below, ~above)
+
+    image[below] = y_min
+    image[above] = y_max
+    if between.any():
+        image[between] = ((image[between] - c) / w + 0.5) * y_range + y_min
+
+    return image