AC: respect calculate_average in multi_accuracy and f1-score

tools/accuracy_checker/accuracy_checker/metrics/multilabel_recognition.py

@@ -134,6 +134,8 @@ def update(self, annotation, prediction):
     def evaluate(self, annotations, predictions):
         tp_tn = np.add(self.tp, self.tn, dtype=float)
         per_class = np.divide(tp_tn, self.counter, out=np.zeros_like(tp_tn, dtype=float), where=self.counter != 0)
+        if not self.calculate_average:
+            return per_class
         average = np.sum(tp_tn) / np.sum(self.counter)
 
         return [*per_class, average]
@@ -230,9 +232,11 @@ def update(self, annotation, prediction):
     def evaluate(self, annotations, predictions):
         precisions = self.precision.evaluate(annotations, predictions)
         recalls = self.recall.evaluate(annotations, predictions)
+        # precision and recall share this config, so they only end with an average if calculate_average is set
+        num_classes = len(self.labels)
 
-        precision_add = np.add(precisions[:-1], recalls[:-1], dtype=float)
-        precision_multiply = np.multiply(precisions[:-1], recalls[:-1], dtype=float)
+        precision_add = np.add(precisions[:num_classes], recalls[:num_classes], dtype=float)
+        precision_multiply = np.multiply(precisions[:num_classes], recalls[:num_classes], dtype=float)
 
         per_class = 2 * np.divide(
             precision_multiply, precision_add, out=np.zeros_like(precision_multiply, dtype=float),

tools/accuracy_checker/tests/test_multilabel_metrics.py

@@ -0,0 +1,61 @@
+"""
+Copyright (c) 2026 Intel Corporation
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+      http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+"""
+
+
+import pytest
+from accuracy_checker.metrics.multilabel_recognition import F1Score, MultiLabelAccuracy
+from accuracy_checker.representation import MultiLabelRecognitionAnnotation, MultiLabelRecognitionPrediction
+from .common import DummyDataset
+
+
+LABEL_MAP = {0: 'a', 1: 'b', 2: 'c'}
+
+
+def evaluate(metric_cls, config):
+    metric = metric_cls(config, DummyDataset(label_map=LABEL_MAP))
+    annotations = [
+        MultiLabelRecognitionAnnotation('img1', [1, 0, 1]),
+        MultiLabelRecognitionAnnotation('img2', [0, 1, 1]),
+    ]
+    predictions = [
+        MultiLabelRecognitionPrediction('img1', [1, 1, 1]),
+        MultiLabelRecognitionPrediction('img2', [0, 1, 0]),
+    ]
+    for annotation, prediction in zip(annotations, predictions):
+        metric.update(annotation, prediction)
+    return list(metric.evaluate(annotations, predictions)), metric.meta['names']
+
+
+class TestMultiLabelMetrics:
+    def test_f1_score_with_average(self):
+        result, names = evaluate(F1Score, {'type': 'f1-score'})
+        assert names == ['a', 'b', 'c', 'average']
+        assert result == pytest.approx([1.0, 2 / 3, 2 / 3, 0.75])
+
+    def test_f1_score_without_average_keeps_all_classes(self):
+        result, names = evaluate(F1Score, {'type': 'f1-score', 'calculate_average': False})
+        assert names == ['a', 'b', 'c']
+        assert result == pytest.approx([1.0, 2 / 3, 2 / 3])
+
+    def test_multi_accuracy_with_average(self):
+        result, names = evaluate(MultiLabelAccuracy, {'type': 'multi_accuracy'})
+        assert names == ['a', 'b', 'c', 'average']
+        assert result == pytest.approx([1.0, 0.5, 0.5, 2 / 3])
+
+    def test_multi_accuracy_without_average(self):
+        result, names = evaluate(MultiLabelAccuracy, {'type': 'multi_accuracy', 'calculate_average': False})
+        assert names == ['a', 'b', 'c']
+        assert result == pytest.approx([1.0, 0.5, 0.5])