added floating point conversion to label mapper for json serializability

image_prediction/label_mapper/mappers/probability.py

@@ -1,7 +1,9 @@
+from functools import partial
 from operator import itemgetter
 from typing import Mapping, Iterable
 
 import numpy as np
+from funcy import rcompose
 
 from image_prediction.exceptions import UnexpectedLabelFormat
 from image_prediction.label_mapper.mapper import LabelMapper
@@ -10,6 +12,9 @@ from image_prediction.label_mapper.mapper import LabelMapper
 class ProbabilityMapper(LabelMapper):
     def __init__(self, labels: Mapping[int, str]):
         self.__labels = labels
+        # String conversion in the middle due to floating point precision issues.
+        # See: https://stackoverflow.com/questions/56820/round-doesnt-seem-to-be-rounding-properly
+        self.__rounder = rcompose(lambda d: round(d, 4), str, float)
 
     def __validate_array_label_format(self, probabilities: np.ndarray) -> None:
         if not len(probabilities) == len(self.__labels):
@@ -19,7 +24,7 @@ class ProbabilityMapper(LabelMapper):
 
     def __map_array(self, probabilities: np.ndarray) -> dict:
         self.__validate_array_label_format(probabilities)
-        cls2prob = dict(sorted(zip(self.__labels, probabilities), key=itemgetter(1), reverse=True))
+        cls2prob = dict(sorted(zip(self.__labels, list(map(self.__rounder, probabilities))), key=itemgetter(1), reverse=True))
         most_likely = [*cls2prob][0]
         return {"label": most_likely, "probabilities": cls2prob}
 

image_prediction/pipeline.py

@@ -9,6 +9,7 @@ from image_prediction.estimator.adapter.adapter import EstimatorAdapter
 from image_prediction.extractor_classifier.extractor_classifier import ExtractorClassifier
 from image_prediction.formatter.formatters.info_formatter import EnumFormatter
 from image_prediction.image_extractor.extractors.parsable import ParsablePDFImageExtractor
+from image_prediction.label_mapper.mappers.probability import ProbabilityMapper
 from image_prediction.locations import MLRUNS_DIR
 from image_prediction.model_loader.loader import ModelLoader
 from image_prediction.model_loader.loaders.mlflow import MlflowConnector
@@ -21,7 +22,8 @@ def get_image_classifier():
     model_loader = ModelLoader(MlflowConnector(MlflowModelReader(MLRUNS_DIR)))
     model = model_loader.load_model(CONFIG.service.run_id)
     classes = model_loader.load_classes(CONFIG.service.run_id)
-    classifier = Classifier(EstimatorAdapter(model), classes)
+    label_mapper = ProbabilityMapper(classes)
+    classifier = Classifier(EstimatorAdapter(model), label_mapper)
     image_classifier = ImageClassifier(classifier)
 
     return image_classifier

scripts/run_pipeline.py

@@ -0,0 +1,29 @@
+import argparse
+import json
+
+from image_prediction.pipeline import Pipeline
+
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("pdf")
+
+    args = parser.parse_args()
+
+    return args
+
+
+def main(args):
+
+    pipeline = Pipeline()
+
+    with open(args.pdf, "rb") as f:
+        predictions = pipeline(f.read())
+
+    for prd in predictions:
+        print(prd)
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

test/unit_tests/conftest.py

@@ -1,6 +1,7 @@
 import random
 import string
 import tempfile
+from functools import partial
 from itertools import starmap
 from operator import itemgetter
 
@@ -8,6 +9,7 @@ import fpdf
 import numpy as np
 import pytest
 from PIL import Image
+from funcy import rcompose
 
 from image_prediction.classifier.classifier import Classifier
 from image_prediction.classifier.image_classifier import ImageClassifier
@@ -196,10 +198,11 @@ def batch_of_expected_numeric_labels(batch_size, classes):
 @pytest.fixture
 def batch_of_expected_label_to_probability_mappings(batch_of_expected_probability_arrays, classes):
     def map_probabilities(probabilities):
-        lbl2prob = dict(sorted(zip(classes, probabilities), key=itemgetter(1), reverse=True))
+        lbl2prob = dict(sorted(zip(classes, map(rounder, probabilities)), key=itemgetter(1), reverse=True))
         most_likely = [*lbl2prob][0]
         return {"label": most_likely, "probabilities": lbl2prob}
 
+    rounder = rcompose(partial(np.round, decimals=4), float)
     return list(map(map_probabilities, batch_of_expected_probability_arrays))