Python vrd_evaluation.label_data_type方法代码示例

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def build_groundtruth_vrd_dictionary(data, class_label_map,
                                     relationship_label_map):
  """Builds a groundtruth dictionary from groundtruth data in CSV file.

  Args:
    data: Pandas DataFrame with the groundtruth data for a single image.
    class_label_map: Class labelmap from string label name to an integer.
    relationship_label_map: Relationship type labelmap from string name to an
      integer.

  Returns:
    A dictionary with keys suitable for passing to
    VRDDetectionEvaluator.add_single_ground_truth_image_info:
        standard_fields.InputDataFields.groundtruth_boxes: A numpy array
          of structures with the shape [M, 1], representing M tuples, each tuple
          containing the same number of named bounding boxes.
          Each box is of the format [y_min, x_min, y_max, x_max] (see
          datatype vrd_box_data_type, single_box_data_type above).
        standard_fields.InputDataFields.groundtruth_classes: A numpy array of
          structures shape [M, 1], representing  the class labels of the
          corresponding bounding boxes and possibly additional classes (see
          datatype label_data_type above).
        standard_fields.InputDataFields.verified_labels: numpy array
          of shape [K] containing verified labels.
  """
  data_boxes = data[data.LabelName.isnull()]
  data_labels = data[data.LabelName1.isnull()]

  boxes = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.vrd_box_data_type)
  boxes['subject'] = data_boxes[['YMin1', 'XMin1', 'YMax1',
                                 'XMax1']].as_matrix()
  boxes['object'] = data_boxes[['YMin2', 'XMin2', 'YMax2', 'XMax2']].as_matrix()

  labels = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.label_data_type)
  labels['subject'] = data_boxes['LabelName1'].map(
      lambda x: class_label_map[x]).as_matrix()
  labels['object'] = data_boxes['LabelName2'].map(
      lambda x: class_label_map[x]).as_matrix()
  labels['relation'] = data_boxes['RelationshipLabel'].map(
      lambda x: relationship_label_map[x]).as_matrix()

  return {
      standard_fields.InputDataFields.groundtruth_boxes:
          boxes,
      standard_fields.InputDataFields.groundtruth_classes:
          labels,
      standard_fields.InputDataFields.groundtruth_image_classes:
          data_labels['LabelName'].map(lambda x: class_label_map[x])
          .as_matrix(),
  } 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def build_predictions_vrd_dictionary(data, class_label_map,
                                     relationship_label_map):
  """Builds a predictions dictionary from predictions data in CSV file.

  Args:
    data: Pandas DataFrame with the predictions data for a single image.
    class_label_map: Class labelmap from string label name to an integer.
    relationship_label_map: Relationship type labelmap from string name to an
      integer.

  Returns:
    Dictionary with keys suitable for passing to
    VRDDetectionEvaluator.add_single_detected_image_info:
        standard_fields.DetectionResultFields.detection_boxes: A numpy array of
          structures with shape [N, 1], representing N tuples, each tuple
          containing the same number of named bounding boxes.
          Each box is of the format [y_min, x_min, y_max, x_max] (as an example
          see datatype vrd_box_data_type, single_box_data_type above).
        standard_fields.DetectionResultFields.detection_scores: float32 numpy
          array of shape [N] containing detection scores for the boxes.
        standard_fields.DetectionResultFields.detection_classes: A numpy array
          of structures shape [N, 1], representing the class labels of the
          corresponding bounding boxes and possibly additional classes (see
          datatype label_data_type above).
  """
  data_boxes = data

  boxes = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.vrd_box_data_type)
  boxes['subject'] = data_boxes[['YMin1', 'XMin1', 'YMax1',
                                 'XMax1']].as_matrix()
  boxes['object'] = data_boxes[['YMin2', 'XMin2', 'YMax2', 'XMax2']].as_matrix()

  labels = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.label_data_type)
  labels['subject'] = data_boxes['LabelName1'].map(
      lambda x: class_label_map[x]).as_matrix()
  labels['object'] = data_boxes['LabelName2'].map(
      lambda x: class_label_map[x]).as_matrix()
  labels['relation'] = data_boxes['RelationshipLabel'].map(
      lambda x: relationship_label_map[x]).as_matrix()

  return {
      standard_fields.DetectionResultFields.detection_boxes:
          boxes,
      standard_fields.DetectionResultFields.detection_classes:
          labels,
      standard_fields.DetectionResultFields.detection_scores:
          data_boxes['Score'].as_matrix()
  } 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def testBuildGroundtruthDictionary(self):
    np_data = pd.DataFrame(
        [[
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.3, 0.5, 0.6,
            0.0, 0.3, 0.5, 0.6, 'is', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/02gy9n', 0.0, 0.3, 0.5, 0.6,
            0.1, 0.2, 0.3, 0.4, 'under', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.1, 0.2, 0.3,
            0.0, 0.1, 0.2, 0.3, 'is', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/083vt', '/m/04bcr3', 0.1, 0.2, 0.3, 0.4,
            0.5, 0.6, 0.7, 0.8, 'at', None, None
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/04bcr3', 1.0
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/083vt', 0.0
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/02gy9n', 0.0
        ]],
        columns=[
            'ImageID', 'LabelName1', 'LabelName2', 'XMin1', 'XMax1', 'YMin1',
            'YMax1', 'XMin2', 'XMax2', 'YMin2', 'YMax2', 'RelationshipLabel',
            'LabelName', 'Confidence'
        ])
    class_label_map = {'/m/04bcr3': 1, '/m/083vt': 2, '/m/02gy9n': 3}
    relationship_label_map = {'is': 1, 'under': 2, 'at': 3}
    groundtruth_dictionary = utils.build_groundtruth_vrd_dictionary(
        np_data, class_label_map, relationship_label_map)

    self.assertTrue(standard_fields.InputDataFields.groundtruth_boxes in
                    groundtruth_dictionary)
    self.assertTrue(standard_fields.InputDataFields.groundtruth_classes in
                    groundtruth_dictionary)
    self.assertTrue(standard_fields.InputDataFields.groundtruth_image_classes in
                    groundtruth_dictionary)

    self.assertAllEqual(
        np.array(
            [(1, 2, 1), (1, 3, 2), (1, 2, 1), (2, 1, 3)],
            dtype=vrd_evaluation.label_data_type), groundtruth_dictionary[
                standard_fields.InputDataFields.groundtruth_classes])
    expected_vrd_data = np.array(
        [
            ([0.5, 0.0, 0.6, 0.3], [0.5, 0.0, 0.6, 0.3]),
            ([0.5, 0.0, 0.6, 0.3], [0.3, 0.1, 0.4, 0.2]),
            ([0.2, 0.0, 0.3, 0.1], [0.2, 0.0, 0.3, 0.1]),
            ([0.3, 0.1, 0.4, 0.2], [0.7, 0.5, 0.8, 0.6]),
        ],
        dtype=vrd_evaluation.vrd_box_data_type)
    for field in expected_vrd_data.dtype.fields:
      self.assertNDArrayNear(
          expected_vrd_data[field], groundtruth_dictionary[
              standard_fields.InputDataFields.groundtruth_boxes][field], 1e-5)
    self.assertAllEqual(
        np.array([1, 2, 3]), groundtruth_dictionary[
            standard_fields.InputDataFields.groundtruth_image_classes]) 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def testBuildPredictionDictionary(self):
    np_data = pd.DataFrame(
        [[
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.3, 0.5, 0.6,
            0.0, 0.3, 0.5, 0.6, 'is', 0.1
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/02gy9n', 0.0, 0.3, 0.5, 0.6,
            0.1, 0.2, 0.3, 0.4, 'under', 0.2
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.1, 0.2, 0.3,
            0.0, 0.1, 0.2, 0.3, 'is', 0.3
        ], [
            'fe58ec1b06db2bb7', '/m/083vt', '/m/04bcr3', 0.1, 0.2, 0.3, 0.4,
            0.5, 0.6, 0.7, 0.8, 'at', 0.4
        ]],
        columns=[
            'ImageID', 'LabelName1', 'LabelName2', 'XMin1', 'XMax1', 'YMin1',
            'YMax1', 'XMin2', 'XMax2', 'YMin2', 'YMax2', 'RelationshipLabel',
            'Score'
        ])
    class_label_map = {'/m/04bcr3': 1, '/m/083vt': 2, '/m/02gy9n': 3}
    relationship_label_map = {'is': 1, 'under': 2, 'at': 3}
    prediction_dictionary = utils.build_predictions_vrd_dictionary(
        np_data, class_label_map, relationship_label_map)

    self.assertTrue(standard_fields.DetectionResultFields.detection_boxes in
                    prediction_dictionary)
    self.assertTrue(standard_fields.DetectionResultFields.detection_classes in
                    prediction_dictionary)
    self.assertTrue(standard_fields.DetectionResultFields.detection_scores in
                    prediction_dictionary)

    self.assertAllEqual(
        np.array(
            [(1, 2, 1), (1, 3, 2), (1, 2, 1), (2, 1, 3)],
            dtype=vrd_evaluation.label_data_type), prediction_dictionary[
                standard_fields.DetectionResultFields.detection_classes])
    expected_vrd_data = np.array(
        [
            ([0.5, 0.0, 0.6, 0.3], [0.5, 0.0, 0.6, 0.3]),
            ([0.5, 0.0, 0.6, 0.3], [0.3, 0.1, 0.4, 0.2]),
            ([0.2, 0.0, 0.3, 0.1], [0.2, 0.0, 0.3, 0.1]),
            ([0.3, 0.1, 0.4, 0.2], [0.7, 0.5, 0.8, 0.6]),
        ],
        dtype=vrd_evaluation.vrd_box_data_type)
    for field in expected_vrd_data.dtype.fields:
      self.assertNDArrayNear(
          expected_vrd_data[field], prediction_dictionary[
              standard_fields.DetectionResultFields.detection_boxes][field],
          1e-5)
    self.assertNDArrayNear(
        np.array([0.1, 0.2, 0.3, 0.4]), prediction_dictionary[
            standard_fields.DetectionResultFields.detection_scores], 1e-5) 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def build_groundtruth_vrd_dictionary(data, class_label_map,
                                     relationship_label_map):
  """Builds a groundtruth dictionary from groundtruth data in CSV file.

  Args:
    data: Pandas DataFrame with the groundtruth data for a single image.
    class_label_map: Class labelmap from string label name to an integer.
    relationship_label_map: Relationship type labelmap from string name to an
      integer.

  Returns:
    A dictionary with keys suitable for passing to
    VRDDetectionEvaluator.add_single_ground_truth_image_info:
        standard_fields.InputDataFields.groundtruth_boxes: A numpy array
          of structures with the shape [M, 1], representing M tuples, each tuple
          containing the same number of named bounding boxes.
          Each box is of the format [y_min, x_min, y_max, x_max] (see
          datatype vrd_box_data_type, single_box_data_type above).
        standard_fields.InputDataFields.groundtruth_classes: A numpy array of
          structures shape [M, 1], representing  the class labels of the
          corresponding bounding boxes and possibly additional classes (see
          datatype label_data_type above).
        standard_fields.InputDataFields.verified_labels: numpy array
          of shape [K] containing verified labels.
  """
  data_boxes = data[data.LabelName.isnull()]
  data_labels = data[data.LabelName1.isnull()]

  boxes = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.vrd_box_data_type)
  boxes['subject'] = data_boxes[['YMin1', 'XMin1', 'YMax1',
                                 'XMax1']].as_matrix()
  boxes['object'] = data_boxes[['YMin2', 'XMin2', 'YMax2', 'XMax2']].as_matrix()

  labels = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.label_data_type)
  labels['subject'] = data_boxes['LabelName1'].map(lambda x: class_label_map[x])
  labels['object'] = data_boxes['LabelName2'].map(lambda x: class_label_map[x])
  labels['relation'] = data_boxes['RelationshipLabel'].map(
      lambda x: relationship_label_map[x])

  return {
      standard_fields.InputDataFields.groundtruth_boxes:
          boxes,
      standard_fields.InputDataFields.groundtruth_classes:
          labels,
      standard_fields.InputDataFields.verified_labels:
          data_labels['LabelName'].map(lambda x: class_label_map[x]),
  } 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def build_predictions_vrd_dictionary(data, class_label_map,
                                     relationship_label_map):
  """Builds a predictions dictionary from predictions data in CSV file.

  Args:
    data: Pandas DataFrame with the predictions data for a single image.
    class_label_map: Class labelmap from string label name to an integer.
    relationship_label_map: Relationship type labelmap from string name to an
      integer.

  Returns:
    Dictionary with keys suitable for passing to
    VRDDetectionEvaluator.add_single_detected_image_info:
        standard_fields.DetectionResultFields.detection_boxes: A numpy array of
          structures with shape [N, 1], representing N tuples, each tuple
          containing the same number of named bounding boxes.
          Each box is of the format [y_min, x_min, y_max, x_max] (as an example
          see datatype vrd_box_data_type, single_box_data_type above).
        standard_fields.DetectionResultFields.detection_scores: float32 numpy
          array of shape [N] containing detection scores for the boxes.
        standard_fields.DetectionResultFields.detection_classes: A numpy array
          of structures shape [N, 1], representing the class labels of the
          corresponding bounding boxes and possibly additional classes (see
          datatype label_data_type above).
  """
  data_boxes = data

  boxes = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.vrd_box_data_type)
  boxes['subject'] = data_boxes[['YMin1', 'XMin1', 'YMax1',
                                 'XMax1']].as_matrix()
  boxes['object'] = data_boxes[['YMin2', 'XMin2', 'YMax2', 'XMax2']].as_matrix()

  labels = np.zeros(data_boxes.shape[0], dtype=vrd_evaluation.label_data_type)
  labels['subject'] = data_boxes['LabelName1'].map(lambda x: class_label_map[x])
  labels['object'] = data_boxes['LabelName2'].map(lambda x: class_label_map[x])
  labels['relation'] = data_boxes['RelationshipLabel'].map(
      lambda x: relationship_label_map[x])

  return {
      standard_fields.DetectionResultFields.detection_boxes:
          boxes,
      standard_fields.DetectionResultFields.detection_classes:
          labels,
      standard_fields.DetectionResultFields.detection_scores:
          data_boxes['Score'].as_matrix()
  } 

# 需要导入模块: from object_detection.utils import vrd_evaluation [as 别名]
# 或者: from object_detection.utils.vrd_evaluation import label_data_type [as 别名]
def testBuildGroundtruthDictionary(self):
    np_data = pd.DataFrame(
        [[
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.3, 0.5, 0.6,
            0.0, 0.3, 0.5, 0.6, 'is', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/02gy9n', 0.0, 0.3, 0.5, 0.6,
            0.1, 0.2, 0.3, 0.4, 'under', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/04bcr3', '/m/083vt', 0.0, 0.1, 0.2, 0.3,
            0.0, 0.1, 0.2, 0.3, 'is', None, None
        ], [
            'fe58ec1b06db2bb7', '/m/083vt', '/m/04bcr3', 0.1, 0.2, 0.3, 0.4,
            0.5, 0.6, 0.7, 0.8, 'at', None, None
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/04bcr3', 1.0
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/083vt', 0.0
        ], [
            'fe58ec1b06db2bb7', None, None, None, None, None, None, None, None,
            None, None, None, '/m/02gy9n', 0.0
        ]],
        columns=[
            'ImageID', 'LabelName1', 'LabelName2', 'XMin1', 'XMax1', 'YMin1',
            'YMax1', 'XMin2', 'XMax2', 'YMin2', 'YMax2', 'RelationshipLabel',
            'LabelName', 'Confidence'
        ])
    class_label_map = {'/m/04bcr3': 1, '/m/083vt': 2, '/m/02gy9n': 3}
    relationship_label_map = {'is': 1, 'under': 2, 'at': 3}
    groundtruth_dictionary = utils.build_groundtruth_vrd_dictionary(
        np_data, class_label_map, relationship_label_map)

    self.assertTrue(standard_fields.InputDataFields.groundtruth_boxes in
                    groundtruth_dictionary)
    self.assertTrue(standard_fields.InputDataFields.groundtruth_classes in
                    groundtruth_dictionary)
    self.assertTrue(standard_fields.InputDataFields.verified_labels in
                    groundtruth_dictionary)

    self.assertAllEqual(
        np.array(
            [(1, 2, 1), (1, 3, 2), (1, 2, 1), (2, 1, 3)],
            dtype=vrd_evaluation.label_data_type), groundtruth_dictionary[
                standard_fields.InputDataFields.groundtruth_classes])
    expected_vrd_data = np.array(
        [
            ([0.5, 0.0, 0.6, 0.3], [0.5, 0.0, 0.6, 0.3]),
            ([0.5, 0.0, 0.6, 0.3], [0.3, 0.1, 0.4, 0.2]),
            ([0.2, 0.0, 0.3, 0.1], [0.2, 0.0, 0.3, 0.1]),
            ([0.3, 0.1, 0.4, 0.2], [0.7, 0.5, 0.8, 0.6]),
        ],
        dtype=vrd_evaluation.vrd_box_data_type)
    for field in expected_vrd_data.dtype.fields:
      self.assertNDArrayNear(
          expected_vrd_data[field], groundtruth_dictionary[
              standard_fields.InputDataFields.groundtruth_boxes][field], 1e-5)
    self.assertAllEqual(
        np.array([1, 2, 3]),
        groundtruth_dictionary[standard_fields.InputDataFields.verified_labels])