Feature Extraction API

class sdc_scissor.feature_extraction_api.feature_extraction.RoadFeatures
__init__()

Class representing the road features as attributes

to_dict()

Serialize the road features to a python dictionary

Returns:

A dictionary with the road features

class sdc_scissor.feature_extraction_api.feature_extraction.RoadSegment
__init__()

A class representing a road segment

class sdc_scissor.feature_extraction_api.feature_extraction.SegmentType
l_turn = 'left_turn'
r_turn = 'right_turn'
straight = 'straight'
class sdc_scissor.feature_extraction_api.feature_extraction.FeatureExtractor(segmentation_strategy)
__init__(segmentation_strategy)

A class that is responsible for extracting road features of test scenarios.

Parameters:

segmentation_strategy – A road segmentation strategy

static save_to_csv(road_features: list, out_dir: Path)

Save the road features to a csv file.

Parameters:

  • road_features – List of road features

  • out_dir – Path to store the csv file

extract_features(test: Test) → RoadFeatures

Input is a list of (x, y) tuples which defines the road. This function extract the angles and radius of segments. Furthermore, the statistics of angles and radius are calculated.

Parameters:

test – A test object

Returns:

A road feature object

class sdc_scissor.feature_extraction_api.road_geometry_calculator.RoadGeometryCalculator
extract_turn_angles(road_points)

Extract angles of raod points and ad them to the instance variable.

Parameters:

road_points – Points that define the road in the test scenario.

Returns:

Angles in degrees of road’s turns defined by the road points.

static get_angle(first_vec, second_vec)

Returns the angle in degrees between the first and second vector. A left turn as positive angles whereas right turns have negatives.

Parameters:

  • first_vec – First 2D vector

  • second_vec – Second 3D vector

Returns:

Angle between the vectors in degrees

static get_distance_between(first_point, second_point) → float

Calculate Euclidean distance between two points.

Parameters:

  • first_point – (x,y) coordinates of first point

  • second_point – (x,y) coordinates of second point

Returns:

Euclidean distance between the points

static get_direction(first_point, second_point)

Get the direction vector from the first point to the second point.

Parameters:

  • first_point – (x,y) coordinates of first point

  • second_point – (x,y) coordinates of second point

Returns:

Return the difference 2D vector (second_point-first_point)

static get_road_length(road_points)

Compute the length of the road.

Parameters:

road_points – List of road points defining the road

Returns:

Length of the road

class sdc_scissor.feature_extraction_api.segmentation_strategy.SegmentationStrategy

Interface for implementing different segmentation strategies.

abstract extract_segments(road_points)

Abstract method to retrieve segments from a road, based on a specific strategy.

Parameters:

road_points – Sequence of coordinates specifying the road.

class sdc_scissor.feature_extraction_api.angle_based_strategy.AngleBasedStrategy(angle_threshold=5, decision_distance=10)

Define segments based on identified turns. The center of a turn where the angles possibly reaches the maximum will also be the center of the segment. Straight segments are identified where no significant turn angles are present.

__init__(angle_threshold=5, decision_distance=10)

Instantiate a strategy object defining the segmentation strategy of the road. Class is coherent to the strategy pattern.

Parameters:

  • angle_threshold – Angle threshold defining a new segment.

  • decision_distance – Road distance for calculating the turn angle.

extract_segments(road_points)

Extract segments from road points according to the angle-based segmentation strategy.

Parameters:

road_points – List of 2D points, defining the road to be segmented

Returns:

List of indexes defining the start and end of segments

class sdc_scissor.feature_extraction_api.equi_distance_strategy.EquiDistanceStrategy(number_of_segments)

Define segments with equal lengts. The last segment might be different to the others since there might be a remainder while dividing the road into predefined number of segments.

__init__(number_of_segments)

Initialize the strategy by providing the number of segments that should be generated

Parameters:

number_of_segments – The number of segments with equal distances that should be generated

extract_segments(road_points)

Extract the segments of the road specified by the road points

Parameters:

road_points – Road points defining the road

Returns:

List of indexes representing the start and end point of segments

class sdc_scissor.feature_extraction_api.parameterized_uniform_strategy.ParameterizedUniformStrategy(risk_factor: str, max_seg_length_to_full_road)

This class segments the road based on parameters like risk factor and fullroad length.

__init__(risk_factor: str, max_seg_length_to_full_road)
Parameters:

  • risk_factor – Risk factor of the driving AI.

  • max_seg_length_to_full_road – Relative size of the longest allowed segment compared to the full road length.

extract_segments(road_points)

Returns a list of tuples that contain start end end indexes of road segments.

Parameters:

road_points – List of coordinates specifying the road.

Returns:

List of start and end indexes defining the start and end road points of segments.