--- a
+++ b/pipeline.py
@@ -0,0 +1,84 @@
+import os
+import pandas as pd
+from tqdm import tqdm
+import torch
+from torchvision import transforms
+from torch.utils.data import DataLoader, Dataset
+from torchvision.datasets.folder import pil_loader
+
+data_cat = ['train', 'valid'] # data categories
+
+def get_study_level_data(study_type):
+    """
+    Returns a dict, with keys 'train' and 'valid' and respective values as study level dataframes, 
+    these dataframes contain three columns 'Path', 'Count', 'Label'
+    Args:
+        study_type (string): one of the seven study type folder names in 'train/valid/test' dataset 
+    """
+    study_data = {}
+    study_label = {'positive': 1, 'negative': 0}
+    for phase in data_cat:
+        BASE_DIR = 'MURA-v1.0/%s/%s/' % (phase, study_type)
+        patients = list(os.walk(BASE_DIR))[0][1] # list of patient folder names
+        study_data[phase] = pd.DataFrame(columns=['Path', 'Count', 'Label'])
+        i = 0
+        for patient in tqdm(patients): # for each patient folder
+            for study in os.listdir(BASE_DIR + patient): # for each study in that patient folder
+                label = study_label[study.split('_')[1]] # get label 0 or 1
+                path = BASE_DIR + patient + '/' + study + '/' # path to this study
+                study_data[phase].loc[i] = [path, len(os.listdir(path)), label] # add new row
+                i+=1
+    return study_data
+
+class ImageDataset(Dataset):
+    """training dataset."""
+
+    def __init__(self, df, transform=None):
+        """
+        Args:
+            df (pd.DataFrame): a pandas DataFrame with image path and labels.
+            transform (callable, optional): Optional transform to be applied
+                on a sample.
+        """
+        self.df = df
+        self.transform = transform
+
+    def __len__(self):
+        return len(self.df)
+
+    def __getitem__(self, idx):
+        study_path = self.df.iloc[idx, 0]
+        count = self.df.iloc[idx, 1]
+        images = []
+        for i in range(count):
+            image = pil_loader(study_path + 'image%s.png' % (i+1))
+            images.append(self.transform(image))
+        images = torch.stack(images)
+        label = self.df.iloc[idx, 2]
+        sample = {'images': images, 'label': label}
+        return sample
+
+def get_dataloaders(data, batch_size=8, study_level=False):
+    '''
+    Returns dataloader pipeline with data augmentation
+    '''
+    data_transforms = {
+        'train': transforms.Compose([
+                transforms.Resize((224, 224)),
+                transforms.RandomHorizontalFlip(),
+                transforms.RandomRotation(10),
+                transforms.ToTensor(),
+                transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 
+        ]),
+        'valid': transforms.Compose([
+            transforms.Resize((224, 224)),
+            transforms.ToTensor(),
+            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+        ]),
+    }
+    image_datasets = {x: ImageDataset(data[x], transform=data_transforms[x]) for x in data_cat}
+    dataloaders = {x: DataLoader(image_datasets[x], batch_size=batch_size, shuffle=True, num_workers=4) for x in data_cat}
+    return dataloaders
+
+if __name__=='main':
+    pass