--- a
+++ b/dsb2018_topcoders/albu/src/utils.py
@@ -0,0 +1,64 @@
+import threading
+import numpy as np
+import cv2
+cv2.setNumThreads(0)
+cv2.ocl.setUseOpenCL(False)
+from sklearn.model_selection import KFold
+from config import Config
+import argparse
+import json
+import pandas as pd
+
+
+def get_folds(data, num):
+    kf = KFold(n_splits=num, shuffle=True, random_state=42)
+    kf.get_n_splits(data)
+    return kf.split(data)
+
+
+def get_csv_folds(ds, fn, holdout=False):
+    df = pd.read_csv(fn)
+    # df = df[df['source'] != 'janowczyk']
+    df = df[['img_id', 'fold', 'cluster', 'source']]
+    folds = []
+    polosa_id = '193ffaa5272d5c421ae02130a64d98ad120ec70e4ed97a72cdcd4801ce93b066'
+    galaxy_ids = ['538b7673d507014d83af238876e03617396b70fe27f525f8205a4a96900fbb8e', 'a102535b0e88374bea4a1cfd9ee7cb3822ff54f4ab2a9845d428ec22f9ee2288', 'cb4df20a83b2f38b394c67f1d9d4aef29f9794d5345da3576318374ec3a11490', 'f29fd9c52e04403cd2c7d43b6fe2479292e53b2f61969d25256d2d2aca7c6a81']
+    all_folds_ids = galaxy_ids + [polosa_id]
+    for it in set(df['fold']):
+        toadd = (df['fold'] != it) | (df['img_id'].isin(all_folds_ids)) | (df['source'] == 'wikimedia')
+
+        val = df[(df['fold'] == it)]['img_id'].values.tolist()
+        train_groups = df[toadd]['cluster'].values
+        train_ids = df[toadd]['img_id'].values
+
+        train = []
+        for i in range(len(train_ids)):
+            rep = 1
+            if train_groups[i] in ['b', 'd', 'e', 'm']:
+                rep = 4
+            elif train_groups[i] in ['c', 'n']:
+                rep = 3
+            if train_ids[i] == polosa_id:
+                rep = 5
+            train.extend([train_ids[i]] * rep)
+        folds.append((ds.get_indexes_by_names(train), ds.get_indexes_by_names(val)))
+    return folds
+
+
+def get_config():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('config_path')
+    parser.add_argument('--fold', dest='fold', default=None,
+                        help='fold to process')
+    args = parser.parse_args()
+    with open(args.config_path, 'r') as f:
+        config = json.load(f)
+        print(config)
+    config['fold'] = args.fold
+    return Config(**config)
+
+def update_config(config, **kwargs):
+    d = config._asdict()
+    d.update(**kwargs)
+    print(d)
+    return Config(**d)
\ No newline at end of file