expected_merged_confidences = np.array(
        [[0.8, 0, 0.1, 0, 0], [0, 0, 0, 0, 0.2]], dtype=np.float32