--- a
+++ b/Region/visualize_regiontrain.py
@@ -0,0 +1,248 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Fri Nov  9 14:43:07 2018
+
+@author: Josefine
+"""
+
+import numpy as np
+import sys
+import os
+sys.path.append(os.path.join('.', '..')) 
+import matplotlib.pyplot as plt
+import matplotlib
+import matplotlib.pylab as pylab
+%matplotlib inline
+#
+a = np.load('Results/region_performance.npz')
+dice2 = a['dice']
+acc = a['acc']
+spec = a['spec']
+sens = a['sens']
+rates = a['rates']
+
+#params = {'legend.fontsize': 'x-large',
+#          'figure.figsize': (15, 5),
+#         'axes.labelsize': 'x-large',
+#         'axes.titlesize':'x-large',
+#         'xtick.labelsize':'x-large',
+#         'ytick.labelsize':'x-large'}
+#pylab.rcParams.update(params)
+
+def epoch_average(a,size):
+    new_a = []
+    i = 0
+    while i < len(a):
+        val = np.mean(a[i:i+size])
+        new_a.append(val)
+        i+=size
+    return new_a
+
+dim = 128
+
+# Load validation curves
+n_test = int(2*dim)
+acc_cor = epoch_average(np.load('Results/train_hist/region/valid_acc_cor.npy'),n_test)
+loss_cor = epoch_average(np.load('Results/train_hist/region/valid_loss_cor.npy'),n_test)
+acc_sag = epoch_average(np.load('Results/train_hist/region/valid_acc_sag.npy'),n_test)
+loss_sag = epoch_average(np.load('Results/train_hist/region/valid_loss_sag.npy'),n_test)
+acc_axial = epoch_average(np.load('Results/train_hist/region/valid_acc_axial.npy'),n_test)
+loss_axial = epoch_average(np.load('Results/train_hist/region/valid_loss_axial.npy'),n_test)
+
+acc_axial_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_acc_axial.npy'),n_test)
+loss_axial_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_loss_axial.npy'),n_test)
+acc_sag_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_acc_sag.npy'),n_test)
+loss_sag_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_loss_sag.npy'),n_test)
+acc_cor_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_acc_cor.npy'),n_test)
+loss_cor_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/valid_loss_cor.npy'),n_test)
+
+acc_cor_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_acc_cor.npy'),n_test)
+loss_cor_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_loss_cor.npy'),n_test)
+acc_sag_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_acc_sag.npy'),n_test)
+loss_sag_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_loss_sag.npy'),n_test)
+acc_axial_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_acc_axial.npy'),n_test)
+loss_axial_noaug = epoch_average(np.load('Results/train_hist/region_noaug/valid_loss_axial.npy'),n_test)
+
+acc_axial_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_acc_axial.npy'),n_test)
+loss_axial_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_loss_axial.npy'),n_test)
+acc_sag_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_acc_sag.npy'),n_test)
+loss_sag_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_loss_sag.npy'),n_test)
+acc_cor_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_acc_cor.npy'),n_test)
+loss_cor_nothing = epoch_average(np.load('Results/train_hist/region_nothing/valid_loss_cor.npy'),n_test)
+
+plt.rcParams.update({'font.size': 13})
+
+plt.figure(figsize=(8*4, 8*3))
+plt.subplot(3,4,3)
+matplotlib.ticker.MultipleLocator(0.01)
+plt.plot(acc_cor, label = 'Both')
+plt.plot(acc_cor_noaug, label = 'Only drop out')
+plt.plot(acc_cor_nodrop, label = 'Only augmentation')
+plt.plot(acc_cor_nothing, label = 'Nothing')
+plt.legend(loc='lower right')
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+axes = plt.gca()
+axes.set_ylim([0.95,1])
+plt.title('Validation accuracy coronal network')
+plt.subplot(3,4,4)
+plt.title('Validation loss coronal network')
+plt.plot(loss_cor, label = 'Both')
+plt.plot(loss_cor_noaug, label = 'Only drop out')
+plt.plot(loss_cor_nodrop, label = 'Only augmentation')
+plt.plot(loss_cor_nothing, label = 'Nothing')
+plt.legend(loc='upper right')
+axes = plt.gca()
+axes.set_ylim([0.02,0.19])
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+
+plt.subplot(3,4,7)
+plt.plot(acc_sag, label = 'Both')
+plt.plot(acc_sag_noaug, label = 'Only drop out')
+plt.plot(acc_sag_nodrop, label = 'Only augmentation')
+plt.plot(acc_sag_nothing, label = 'Nothing')
+axes = plt.gca()
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+axes.set_ylim([0.95,1])
+plt.legend(loc='lower right')
+plt.title('Validation accuracy sagittal network')
+plt.subplot(3,4,8)
+plt.title('Validation loss sigittal network')
+plt.plot(loss_sag, label = 'Both')
+plt.plot(loss_sag_noaug, label = 'Only drop out')
+plt.plot(loss_sag_nodrop, label = 'Only augmentation')
+plt.plot(loss_sag_nothing, label = 'Nothing')
+plt.legend(loc='upper right')
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+axes = plt.gca()
+axes.set_ylim([0.02,0.19])
+
+plt.subplot(3,4,11)
+plt.plot(acc_axial, label = 'Both')
+plt.plot(acc_axial_noaug, label = 'Only drop out')
+plt.plot(acc_axial_nodrop, label = 'Only augmentation')
+plt.plot(acc_axial_nothing, label = 'Nothing')
+axes = plt.gca()
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+axes.set_ylim([0.95,1])
+plt.title('Validation accuracy axial network')
+plt.legend(loc='lower right')
+plt.subplot(3,4,12)
+plt.plot(loss_axial, label = 'Both')
+plt.plot(loss_axial_noaug, label = 'Only drop out')
+plt.plot(loss_axial_nodrop, label = 'Only augmentation')
+plt.plot(loss_axial_nothing, label = 'Nothing')
+plt.title('Validation loss axial network')
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+plt.legend(loc='upper right')
+axes = plt.gca()
+axes.set_ylim([0.02,0.19])
+
+
+#% Load training curves
+n_test = int(13*6*dim)
+acc_cor = epoch_average(np.load('Results/train_hist/region/train_acc_cor.npy'),n_test)
+loss_cor = epoch_average(np.load('Results/train_hist/region/train_loss_cor.npy'),n_test)
+acc_sag = epoch_average(np.load('Results/train_hist/region/train_acc_sag.npy'),n_test)
+loss_sag = epoch_average(np.load('Results/train_hist/region/train_loss_sag.npy'),n_test)
+acc_axial = epoch_average(np.load('Results/train_hist/region/train_acc_axial.npy'),n_test)
+loss_axial = epoch_average(np.load('Results/train_hist/region/train_loss_axial.npy'),n_test)
+
+n_test = int(13*6*dim)
+acc_axial_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_acc_axial.npy'),n_test)
+loss_axial_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_loss_axial.npy'),n_test)
+acc_sag_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_acc_sag.npy'),n_test)
+loss_sag_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_loss_sag.npy'),n_test)
+acc_cor_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_acc_cor.npy'),n_test)
+loss_cor_nodrop = epoch_average(np.load('Results/train_hist/region_nodrop/train_loss_cor.npy'),n_test)
+
+n_test = int(13*dim)
+acc_cor_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_acc_cor.npy'),n_test)
+loss_cor_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_loss_cor.npy'),n_test)
+acc_sag_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_acc_sag.npy'),n_test)
+loss_sag_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_loss_sag.npy'),n_test)
+acc_axial_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_acc_axial.npy'),n_test)
+loss_axial_noaug = epoch_average(np.load('Results/train_hist/region_noaug/train_loss_axial.npy'),n_test)
+
+acc_axial_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_acc_axial.npy'),n_test)
+loss_axial_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_loss_axial.npy'),n_test)
+acc_sag_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_acc_sag.npy'),n_test)
+loss_sag_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_loss_sag.npy'),n_test)
+acc_cor_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_acc_cor.npy'),n_test)
+loss_cor_nothing = epoch_average(np.load('Results/train_hist/region_nothing/train_loss_cor.npy'),n_test)
+
+plt.subplot(3,4,1)
+plt.plot(acc_cor, label = 'Both')
+plt.plot(acc_cor_noaug, label = 'Only drop out')
+plt.plot(acc_cor_nodrop, label = 'Only augmentation')
+plt.plot(acc_cor_nothing, label = 'Nothing')
+axes = plt.gca()
+axes.set_ylim([0.95,1])
+plt.legend(loc='lower right')
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+plt.title('Train accuracy coronal network')
+plt.subplot(3,4,2)
+plt.title('Train loss coronal network')
+plt.plot(loss_cor, label = 'Both')
+plt.plot(loss_cor_noaug, label = 'Only drop out')
+plt.plot(loss_cor_nodrop, label = 'Only augmentation')
+plt.plot(loss_cor_nothing, label = 'Nothing')
+plt.legend(loc='upper right')
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+axes = plt.gca()
+axes.set_ylim([0,0.05])
+
+plt.subplot(3,4,5)
+plt.plot(acc_sag, label = 'Both')
+plt.plot(acc_sag_noaug, label = 'Only drop out')
+plt.plot(acc_sag_nodrop, label = 'Only augmentation')
+plt.plot(acc_sag_nothing, label = 'Nothing')
+plt.legend(loc='lower right')
+axes = plt.gca()
+axes.set_ylim([0.95,1])
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+plt.title('Train accuracy sagittal network')
+plt.subplot(3,4,6)
+plt.title('Train loss sigittal network')
+plt.plot(loss_sag, label = 'Both')
+plt.plot(loss_sag_noaug, label = 'Only drop out')
+plt.plot(loss_sag_nodrop, label = 'Only augmentation')
+plt.plot(loss_sag_nothing, label = 'Nothing')
+plt.legend(loc='upper right')
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+axes = plt.gca()
+axes.set_ylim([0,0.05])
+
+plt.subplot(3,4,9)
+plt.plot(acc_axial, label = 'Both')
+plt.plot(acc_axial_noaug, label = 'Only drop out')
+plt.plot(acc_axial_nodrop, label = 'Only augmentation')
+plt.plot(acc_axial_nothing, label = 'Nothing')
+plt.legend(loc='lower right')
+axes = plt.gca()
+axes.set_ylim([0.95,1])
+plt.xlabel('Epochs')
+plt.ylabel('Accuracy')
+plt.title('Train accuracy axial network')
+plt.subplot(3,4,10)
+plt.plot(loss_axial, label = 'Both')
+plt.plot(loss_axial_noaug, label = 'Only drop out')
+plt.plot(loss_axial_nodrop, label = 'Only augmentation')
+plt.plot(loss_axial_nothing, label = 'Nothing')
+plt.title('Train loss axial network')
+plt.legend(loc='upper right')
+plt.xlabel('Epochs')
+plt.ylabel('Loss')
+axes = plt.gca()
+axes.set_ylim([0,0.05])
+