Create utils.py

src/ml_things/utils.py

@@ -0,0 +1,100 @@
+"""
+Put random functions I create that need implemented and added to package.
+"""
+
+def plot_figure_image(images_details, title, save_name_path, magnify=3.5, show=True, use_dpi=None):
+  """
+  Plot multiple image side-by-side on a horizontal axis.
+  Great to compare target-prediction.
+
+  Args:
+      images_details (list): List of dictionary that contains images details.
+          Here are arguments for dictionary image details:
+          'image' (numpy): image to plot.
+          'title' (str): Title of the image.
+      
+      title (str): Title of entire figure. It can be large.
+
+      save_name_path (str): Path where to save figure + name of figure.
+
+      magnify (float): Multiply weights and height by magnify value. 
+          Helps zoom in or out figure.
+
+      show (bool): To show the plot or not. If use in headless server no need to show plot.
+
+      use_dpi (int): The dpi of the figure plot. The larger the higher the quality.
+  """
+  import matplotlib.pyplot as plt
+  # Get figure and axes array.
+  figure, axarr = plt.subplots(1,len(images_details), )
+  # For each ax array add sybplot.
+  for index, image_details in enumerate(images_details):
+    # Add subplot image.
+    axarr[index].imshow(image_details['image'])
+    # Add title to subplot.
+    axarr[index].set_title(image_details['title'], fontsize=16)
+  # Add figure title.
+  figure.suptitle(title, fontsize=16, horizontalalignment='left', y=0.75)
+  # figure.subplots_adjust(top=1.0)
+  figure.tight_layout()
+  # Get size of figure.
+  figsize = figure.get_size_inches()
+  # Change size depending on magnify.
+  figsize = [figsize[0] * magnify, figsize[1] * magnify]
+  figure.set_size_inches(figsize)
+  # Set the new figure size.
+  figure.savefig(save_name_path, bbox_inches='tight', pad_inches=0.1, dpi=use_dpi)
+  # Show plot.
+  plt.show() if show else None #print("Not showing figure. It is saved in '{}'.")
+  # Close figure.
+  plt.close(figure)
+
+
+def evaluate_classification(target, prediction):
+  """
+  Compute metrics for classificaiton evlauation.
+
+  Great source: https://datascience.stackexchange.com/a/26855
+  """
+
+  # Make sure confusion matrix is imported.
+  from sklearn.metrics import confusion_matrix
+  # Run confusion matrix over target and prediction
+  tn, fp, fn, tp = confusion_matrix(y_true=target, y_pred=prediction).ravel()
+
+  # Compute various metrics
+  # Precision: (or Positive predictive value)
+  # proportion of predicted positives which are actual positive
+  ppv = tp/(tp+fp)
+  # Sensitivity, hit rate, recall, or true positive rate
+  # proportion of actual positives which are predicted positive
+  tpr = tp/(tp+fn)
+  # Specificity or true negative rate
+  # proportion of actual negative which are predicted negative
+  tnr = tn/(tn+fp)
+  # Negative predictive value
+  npv = tn/(tn+fn)
+  # Fall out or false positive rate
+  fpr = fp/(fp+tn)
+  # False negative rate
+  fnr = fn/(tp+fn)
+  # False discovery rate
+  fdr = fp/(tp+fp)
+  # F1 score
+  f1 = (2*tp)/(2*tp + fp + fn)
+  # Overall accuracy
+  acc = (tp+tn)/(tp+fp+fn+tn)
+  # BCR: Balanced Classification Rate
+  bcr = 0.5 * (tp / (tp + fn) + tn / (tn + fp))
+  # Balanced Error Rate, or HTER
+  ber = 1 - 0.5 * (tp / (tp + fn) + tn / (tn + fp))
+
+  # Return metrics as a dictionary.
+  return {
+          "accuracy": acc,
+          "precision": ppv,
+          "recall": tpr,
+          "f1": f1,
+          "specificity": tnr,
+          'ber': ber,
+          }