Upload data/evaluation/evaluate.py with huggingface_hub

data/evaluation/evaluate.py

@@ -0,0 +1,118 @@
+"""
+Evaluation script for CAMELYON17.
+"""
+
+import sklearn.metrics
+import pandas as ps
+
+import argparse
+
+#----------------------------------------------------------------------------------------------------
+
+def calculate_kappa(reference, submission):
+    """
+    Calculate inter-annotator agreement with quadratic weighted Kappa.
+
+    Args:
+        reference (pandas.DataFrame): List of labels assigned by the organizers.
+        submission (pandas.DataFrame): List of labels assigned by participant.
+
+    Returns:
+        float: Kappa score.
+
+    Raises:
+        ValueError: Unknown stage in reference.
+        ValueError: Patient missing from submission.
+        ValueError: Unknown stage in submission.
+    """
+
+    # The accepted stages are pN0, pN0(i+), pN1mi, pN1, pN2 as described on the website. During parsing all strings converted to lowercase.
+    #
+    stage_list = ['pn0', 'pn0(i+)', 'pn1mi', 'pn1', 'pn2']
+
+    # Extract the patient pN stages from the tables for evaluation.
+    #
+    reference_map = {df_row['patient']: df_row['stage'].lower() for _, df_row in reference.iterrows() if df_row['patient'].lower().endswith('.zip')}
+    submission_map = {df_row['patient']: df_row['stage'].lower() for _, df_row in submission.iterrows() if df_row['patient'].lower().endswith('.zip')}
+
+    # Reorganize data into lists with the same patient order and check consistency.
+    #
+    reference_stage_list = []
+    submission_stage_list = []
+    for patient_id, reference_stage in reference_map.items():
+        # Check consistency: all stages must be from the official stage list and there must be a submission for each patient in the ground truth.
+        #
+        submission_stage = submission_map[patient_id].lower()
+
+        if reference_stage not in stage_list:
+            raise ValueError('Unknown stage in reference: \'{stage}\''.format(stage=reference_stage))
+        if patient_id not in submission_map:
+            raise ValueError('Patient missing from submission: \'{patient}\''.format(patient=patient_id))
+        if submission_stage not in stage_list:
+            raise ValueError('Unknown stage in submission: \'{stage}\''.format(stage=submission_map[patient_id]))
+
+        # Add the pair to the lists.
+        #
+        reference_stage_list.append(reference_stage)
+        submission_stage_list.append(submission_stage)
+
+    # Return the Kappa score.
+    #
+    return sklearn.metrics.cohen_kappa_score(y1=reference_stage_list, y2=submission_stage_list, labels=stage_list, weights='quadratic')
+
+#----------------------------------------------------------------------------------------------------
+
+def collect_arguments():
+    """
+    Collect command line arguments.
+
+    Returns:
+        (str, str): The parsed reference and submission CSV file paths.
+    """
+
+    # Configure argument parser.
+    #
+    argument_parser = argparse.ArgumentParser(description='Calculate inter-annotator agreement.')
+
+    argument_parser.add_argument('-r', '--reference',  required=True, type=str, help='reference CSV path')
+    argument_parser.add_argument('-s', '--submission', required=True, type=str, help='submission CSV path')
+
+    # Parse arguments.
+    #
+    arguments = vars(argument_parser.parse_args())
+
+    # Collect arguments.
+    #
+    parsed_reference_path = arguments['reference']
+    parsed_submission_path = arguments['submission']
+
+    # Print parsed parameters.
+    #
+    print(argument_parser.description)
+    print('Reference: {path}'.format(path=parsed_reference_path))
+    print('Submission: {path}'.format(path=parsed_submission_path))
+
+    return parsed_reference_path, parsed_submission_path
+
+#----------------------------------------------------------------------------------------------------
+
+
+if __name__ == '__main__':
+
+    # Parse parameters.
+    #
+    reference_path, submission_path = collect_arguments()
+
+    # Load tables to Pandas data frames.
+    #
+    reference_df = ps.read_csv(reference_path)
+    submission_df = ps.read_csv(submission_path)
+
+    # Calculate kappa score.
+    #
+    try:
+        kappa_score = calculate_kappa(reference=reference_df, submission=submission_df)
+    except Exception as exception:
+        print(exception)
+    else:
+        print('Score: {score}'.format(score=kappa_score))