Class DescriptorModel

Inheritance Relationships

Derived Types

Class Documentation

class DescriptorModel

The DescriptorModel class is an abstract base class for different types of descriptor models. A descriptor model is a model that can predict a descriptor for a given pose and point, and can be updated with new descriptors.

Subclassed by slam_mer::utils::Average, slam_mer::utils::LastSeen, slam_mer::utils::Medoid, slam_mer::utils::WeightedAverage

Public Types

enum class Type

Types of descriptor models.

Values:

enumerator LAST_SEEN

Descriptor is the last one that was added to the model.

enumerator MEDOID

Descriptor is the medoid of the added descriptors.

enumerator AVERAGE

Descriptor is the arithmetic average of the added descriptors.

enumerator WEIGHTED_AVERAGE

Descriptor is the weighted average of the added descriptors.

typedef std::shared_ptr<DescriptorModel> shared_ptr

Typedef for shared pointer of type DescriptorModel.

Public Functions

virtual ~DescriptorModel() = default

Virtual destructor for the DescriptorModel class.

virtual output_t<descriptor_t> predict(const pose_t &pose_, const point3d_t &point_) = 0

Predicts a descriptor for the provided pose and point.

Parameters:

  • pose_ – Camera pose used by the model.

  • point_ – 3D point used by the model.

Returns:

Predicted descriptor or an error state.

virtual void fit(const descriptor_t &descriptor_, const pose_t &pose_ = pose_t(), const point3d_t &point_ = point3d_t()) = 0

Updates the model with one descriptor observation.

Parameters:

  • descriptor_ – Descriptor measurement to ingest.

  • pose_ – Camera pose associated with the observation.

  • point_ – 3D point associated with the observation.

Public Static Functions

static DescriptorModel::shared_ptr create(const Type &type_)

Factory method that creates a descriptor model instance.

Parameters:

type_ – Descriptor-model type to instantiate.

Returns:

Shared pointer to the requested descriptor-model implementation.

Protected Functions

inline void __point_sanity_check(const point3d_t &point_)

Validates the input point and throws if it is unset.

Parameters:

point_ – Point to validate.

inline void __pose_sanity_check(const pose_t &pose_)

Validates the input pose and throws if it is unset.

Parameters:

pose_ – Pose to validate.

Protected Attributes

descriptor_t _descriptor

The current descriptor of the model. This is updated by the fit method and returned by the predict method.

size_t _numObservations = 0

The number of observations that have been added to the model.