open3d::ml::contrib Namespace Reference

Data Structures
struct	PointCloud
class	PointXYZ
class	SampledData

Functions
PointXYZ	max_point (std::vector< PointXYZ > points)
PointXYZ	min_point (std::vector< PointXYZ > points)
PointXYZ	operator+ (const PointXYZ A, const PointXYZ B)
PointXYZ	operator- (const PointXYZ A, const PointXYZ B)
PointXYZ	operator* (const PointXYZ P, const float a)
PointXYZ	operator* (const float a, const PointXYZ P)
std::ostream &	operator<< (std::ostream &os, const PointXYZ P)
bool	operator== (const PointXYZ A, const PointXYZ B)
const core::Tensor	KnnSearch (const core::Tensor &query_points, const core::Tensor &dataset_points, int knn)
const core::Tensor	RadiusSearch (const core::Tensor &query_points, const core::Tensor &dataset_points, const core::Tensor &query_batches, const core::Tensor &dataset_batches, double radius)
void	grid_subsampling (std::vector< PointXYZ > &original_points, std::vector< PointXYZ > &subsampled_points, std::vector< float > &original_features, std::vector< float > &subsampled_features, std::vector< int > &original_classes, std::vector< int > &subsampled_classes, float sampleDl, int verbose)
void	batch_grid_subsampling (std::vector< PointXYZ > &original_points, std::vector< PointXYZ > &subsampled_points, std::vector< float > &original_features, std::vector< float > &subsampled_features, std::vector< int > &original_classes, std::vector< int > &subsampled_classes, std::vector< int > &original_batches, std::vector< int > &subsampled_batches, float sampleDl, int max_p)
void	brute_neighbors (std::vector< PointXYZ > &queries, std::vector< PointXYZ > &supports, std::vector< int > &neighbors_indices, float radius, int verbose)
void	ordered_neighbors (std::vector< PointXYZ > &queries, std::vector< PointXYZ > &supports, std::vector< int > &neighbors_indices, float radius)
void	batch_nanoflann_neighbors (std::vector< PointXYZ > &queries, std::vector< PointXYZ > &supports, std::vector< int > &q_batches, std::vector< int > &s_batches, std::vector< int > &neighbors_indices, float radius)

Function Documentation

batch_grid_subsampling()

void open3d::ml::contrib::batch_grid_subsampling	(	std::vector< PointXYZ > &	original_points,
		std::vector< PointXYZ > &	subsampled_points,
		std::vector< float > &	original_features,
		std::vector< float > &	subsampled_features,
		std::vector< int > &	original_classes,
		std::vector< int > &	subsampled_classes,
		std::vector< int > &	original_batches,
		std::vector< int > &	subsampled_batches,
		float	sampleDl,
		int	max_p
	)

batch_nanoflann_neighbors()

void open3d::ml::contrib::batch_nanoflann_neighbors	(	std::vector< PointXYZ > &	queries,
		std::vector< PointXYZ > &	supports,
		std::vector< int > &	q_batches,
		std::vector< int > &	s_batches,
		std::vector< int > &	neighbors_indices,
		float	radius
	)

TOOD: This is a temporary function for 3DML repositiory use. In the future, the native Open3D Python API should be improved and used.

Nearest neighbours withing a radius with batching. queries and supports are sliced with their respective batch elements. Uses nanoflann to build a KDTree and find neighbors.

brute_neighbors()

void open3d::ml::contrib::brute_neighbors	(	std::vector< PointXYZ > &	queries,
		std::vector< PointXYZ > &	supports,
		std::vector< int > &	neighbors_indices,
		float	radius,
		int	verbose
	)

grid_subsampling()

void open3d::ml::contrib::grid_subsampling	(	std::vector< PointXYZ > &	original_points,
		std::vector< PointXYZ > &	subsampled_points,
		std::vector< float > &	original_features,
		std::vector< float > &	subsampled_features,
		std::vector< int > &	original_classes,
		std::vector< int > &	subsampled_classes,
		float	sampleDl,
		int	verbose
	)

KnnSearch()

const core::Tensor open3d::ml::contrib::KnnSearch	(	const core::Tensor &	query_points,
		const core::Tensor &	dataset_points,
		int	knn
	)

TOOD: This is a temory wrapper for 3DML repositiory use. In the future, the native Open3D Python API should be improved and used.

Parameters

query_points	Tensor of shape {n_query_points, d}, dtype Float32.
dataset_points	Tensor of shape {n_dataset_points, d}, dtype Float32.
knn	Int.

Returns

Tensor of shape (n_query_points, knn), dtype Int32.

max_point()

PointXYZ open3d::ml::contrib::max_point

(

std::vector< PointXYZ >

points

)

min_point()

PointXYZ open3d::ml::contrib::min_point

(

std::vector< PointXYZ >

points

)

operator*() [1/2]

PointXYZ open3d::ml::contrib::operator* ( const PointXYZ  P, const float  a  )

inline

operator*() [2/2]

PointXYZ open3d::ml::contrib::operator* ( const float  a, const PointXYZ  P  )

inline

operator+()

PointXYZ open3d::ml::contrib::operator+ ( const PointXYZ  A, const PointXYZ  B  )

inline

operator-()

PointXYZ open3d::ml::contrib::operator- ( const PointXYZ  A, const PointXYZ  B  )

inline

operator<<()

std::ostream& open3d::ml::contrib::operator<< ( std::ostream &  os, const PointXYZ  P  )

inline

operator==()

bool open3d::ml::contrib::operator== ( const PointXYZ  A, const PointXYZ  B  )

inline

ordered_neighbors()

void open3d::ml::contrib::ordered_neighbors	(	std::vector< PointXYZ > &	queries,
		std::vector< PointXYZ > &	supports,
		std::vector< int > &	neighbors_indices,
		float	radius
	)

TOOD: This is a temporary function for 3DML repositiory use. In the future, the native Open3D Python API should be improved and used.

Nearest neighbours within a given radius. For each query point, finds a set of neighbor indices whose distance is less than given radius. Modifies the neighbors_indices inplace.

RadiusSearch()

const core::Tensor open3d::ml::contrib::RadiusSearch	(	const core::Tensor &	query_points,
		const core::Tensor &	dataset_points,
		const core::Tensor &	query_batches,
		const core::Tensor &	dataset_batches,
		double	radius
	)

TOOD: This is a temory wrapper for 3DML repositiory use. In the future, the native Open3D Python API should be improved and used.

Parameters

query_points	Tensor of shape {n_query_points, d}, dtype Float32.
dataset_points	Tensor of shape {n_dataset_points, d}, dtype Float32.
query_batches	Tensor of shape {n_batches,}, dtype Int32. It is required that sum(query_batches) == n_query_points.
dataset_batches	Tensor of shape {n_batches,}, dtype Int32. It is required that that sum(dataset_batches) == n_dataset_points.
radius	The radius to search.

Returns

Tensor of shape {n_query_points, max_neighbor}, dtype Int32, where max_neighbor is the maximum number neighbor of neighbors for all query points. For query points with less than max_neighbor neighbors, the neighbor index will be padded by -1.