NetDEM v1.0
Loading...
Searching...
No Matches
netdem::CodedNetSDF Class Reference

A class for representing a shape using a neural network-generated signed distance function and a latent code. More...

#include <shape_coded_netsdf.hpp>

Inheritance diagram for netdem::CodedNetSDF:
netdem::Shape

Public Member Functions

 CodedNetSDF ()
 Default constructor.
 
void InitFromModelFiles (std::string const &net_sdf_file, std::string const &latent_code_file, int id)
 Initialize the CodedNetSDF object from model files.
 
void Init ()
 Initialize the CodedNetSDF object.
 
nlohmann::json PackJson () override
 Pack the CodedNetSDF object into a JSON structure.
 
void InitFromJson (nlohmann::json const &js) override
 Initialize the CodedNetSDF object from a JSON structure.
 
void UpdateSurfaceNodes () override
 Update the surface nodes of the CodedNetSDF object.
 
void UpdateShapeProperties () override
 Update the shape properties of the CodedNetSDF object.
 
void SetSize (double d) override
 Set the size of the CodedNetSDF object.
 
ShapeClone () const override
 Clone the CodedNetSDF object.
 
STLModel GetSTLModel (int num_nodes=200) override
 Generate an STL model of the CodedNetSDF object.
 
double SignedDistance (Vec3d const &pos) const override
 Compute the signed distance from a given position to the surface of the CodedNetSDF object.
 
Vec3d SurfacePoint (Vec3d const &pos) override
 Compute the surface point closest to a given position on the CodedNetSDF object.
 
 ~CodedNetSDF ()
 Destructor.
 
void Print () override
 Print information about the CodedNetSDF object.
 
- Public Member Functions inherited from netdem::Shape
virtual void InitFromJsonFile (std::string const &js_file)
 Initializes the shape instance from a JSON file.
 
virtual double GetSize () const
 Return shape size, which is defined as the diameter of equal-volume sphere.
 
virtual double GetVolume () const
 Return the volume of the shape.
 
virtual Mat3d const & GetInertia () const
 Return the inertia of the shape.
 
Vec3d GetInertiaPrincipal () const
 Return the principal inertia of the shape.
 
virtual double GetBoundSphereRadius () const
 Return the inertia of the shape.
 
virtual std::tuple< Vec3d, Vec3dGetBoundAABB () const
 Gets the boundary AABB of the shape instance.
 
virtual std::tuple< Vec3d, Vec3dGetBoundAABB (Vec3d const &pos, Vec4d const &quat) const
 Gets the boundary AABB of the shape instance.
 
virtual double GetSkin () const
 Return skin size of the shape.
 
virtual double GetSkinFactor () const
 Return skin factor of the shape.
 
virtual double GetSurfaceNodeNum () const
 Return number of surface nodes of the shape.
 
virtual VecXT< Vec3d > const & GetSurfaceNodes () const
 Return surface nodes of the shape.
 
virtual VecXT< double > const & GetSurfaceNodeAreas () const
 Return surface area associated with each surface node.
 
virtual double GetSurfaceNodeSpacing () const
 Return spacing of surface nodes.
 
const STLModelGetRenderMesh () const
 Return the surface mesh for rendering.
 
bool CheckConvexity ()
 Return the convexity of the shape.
 
virtual void Translate (Vec3d const &pos)
 Translates the shape instance to a new position.
 
virtual void UpdateRenderMesh (int num_nodes=200)
 Initializes the surface mesh for rendering.
 
virtual void SetSkin (double sk)
 Set skin size of the shape.
 
virtual void SetSkinFactor (double sk_factor)
 Set skin size factor (skin = size by skin size factor) of the shape.
 
virtual void EnableSurfaceNodes ()
 Enable the use of surface nodes.
 
virtual void DisableSurfaceNodes ()
 Disable the use of surface nodes.
 
virtual bool IsSurfaceNodesEnabled ()
 Check if the use of surface nodes is enabled.
 
virtual void SetSurfaceNodeNum (int num)
 Set number of surface nodes of the shape.
 
virtual void SetSurfaceNodes (VecXT< Vec3d > const &nodes)
 Set the surface nodes of the shape.
 
virtual void SetSurfaceNodes (STLModel const &surf_stl)
 Set the surface nodes of the shape.
 
virtual void SaveAsVTK (std::string const &filename)
 Saves the shape instance as a VTK file.
 
virtual void SaveAsSTL (std::string const &filename)
 Saves the shape instance as an STL file.
 
virtual Vec3d SupportPoint (Vec3d const &dir)
 Gets the support point in a given direction for the shape instance.
 
virtual VecXT< Vec3dSupportPoints (Vec3d const &dir)
 Gets the set of support points in a given direction for the shape instance.
 
virtual bool CheckEnclose (Vec3d const &pos) const
 Determines whether a given point is enclosed within the shape instance.
 
virtual bool CheckPrincipal () const
 Determines whether the shape is in principal configuration.
 
virtual ~Shape ()
 Destructor for the Shape class.
 

Public Attributes

RegressionNet net_sdf_model
 The neural network model used to generate the signed distance function.
 
VecXT< double > latent_code
 The latent code used as input to the neural network model.
 
- Public Attributes inherited from netdem::Shape
int id {0}
 ID of the shape instance.
 
std::string label {"default"}
 Label of the shape instance.
 
Type shape_type {Type::none}
 Type of the shape instance.
 
std::string shape_name {"shape"}
 Name of the shape instance.
 
bool use_customized_solver {false}
 Determines whether to use a customized contact solver or not.
 
STLModel render_mesh
 Surface mesh for rendering.
 

Additional Inherited Members

- Public Types inherited from netdem::Shape
enum  Type {
  none , point_sphere , sphere , spherical_harmonics ,
  trimesh , trimesh_convex , ellipsoid , polybezier ,
  triangle , plane , cylinder , poly_super_ellipsoid ,
  poly_super_quadrics , level_set , netsdf , coded_netsdf ,
  num_shapes
}
 Enumeration for different types of shapes. More...
 
- Static Public Member Functions inherited from netdem::Shape
static MatNd< 8, 3 > VerticesOfAABB (Vec3d const &bmin, Vec3d const &bmax)
 Gets the vertices of the boundary AABB of the shape instance.
 
- Protected Attributes inherited from netdem::Shape
double size {1.0}
 Size and volume of the shape instance.
 
double volume {0.5236}
 Volume of the shape instance.
 
Mat3d inertia {{{0.05236, 0, 0}, {0, 0.05236, 0}, {0, 0, 0.05236}}}
 Inertia tensor of the shape instance.
 
double bound_sphere_radius {0.5}
 Radius of the bounding sphere used for broad-phase contact detection.
 
double skin {0.025}
 Skin thickness used for broad-phase contact detection.
 
double skin_factor {0.05}
 Factor used to adjust the skin thickness.
 
Vec3d bound_aabb_min {-0.5, -0.5, -0.5}
 Minimum bounds of the AABB used for broad-phase contact detection.
 
Vec3d bound_aabb_max {0.5, 0.5, 0.5}
 Maximum bounds of the AABB used for broad-phase contact detection.
 
bool is_convex {true}
 Determines whether the shape instance is convex or not.
 
bool use_surface_nodes {false}
 Determines whether to use nodes for the contact solver.
 
int surface_node_num {1000}
 Number of nodes used for the contact solver.
 
double surface_node_spacing {std::sqrt(Math::PI / 1000.0)}
 Spacing between nodes used for the contact solver.
 
VecXT< Vec3dsurface_nodes
 Surface nodes of the shape instance used for the contact solver.
 
VecXT< double > surface_node_areas
 Area associated with each surface node.
 

Detailed Description

A class for representing a shape using a neural network-generated signed distance function and a latent code.

This class provides methods for loading and initializing a neural network-generated signed distance function and a corresponding latent code, as well as computing signed distances and surface points from a given position.

Examples
90_random_packing.cpp.

Constructor & Destructor Documentation

◆ CodedNetSDF()

CodedNetSDF::CodedNetSDF ( )

Default constructor.

◆ ~CodedNetSDF()

CodedNetSDF::~CodedNetSDF ( )

Destructor.

Member Function Documentation

◆ Clone()

Shape * CodedNetSDF::Clone ( ) const
overridevirtual

Clone the CodedNetSDF object.

This method creates a copy of the CodedNetSDF object.

Returns
A pointer to the cloned CodedNetSDF object.

Reimplemented from netdem::Shape.

◆ GetSTLModel()

STLModel CodedNetSDF::GetSTLModel ( int num_nodes = 200)
overridevirtual

Generate an STL model of the CodedNetSDF object.

This method generates an STL model of the CodedNetSDF object using the specified number of nodes for each surface triangle.

Parameters
num_nodesThe number of nodes to use for each surface triangle (default: 200).
Returns
An STLModel object representing the CodedNetSDF object.

Reimplemented from netdem::Shape.

◆ Init()

void CodedNetSDF::Init ( )

Initialize the CodedNetSDF object.

This method initializes the CodedNetSDF object by loading the neural network model and corresponding latent code from memory.

◆ InitFromJson()

void CodedNetSDF::InitFromJson ( nlohmann::json const & js)
overridevirtual

Initialize the CodedNetSDF object from a JSON structure.

This method initializes the CodedNetSDF object from a JSON structure previously generated by PackJson().

Parameters
jsThe JSON structure to load.

Reimplemented from netdem::Shape.

◆ InitFromModelFiles()

void CodedNetSDF::InitFromModelFiles ( std::string const & net_sdf_file,
std::string const & latent_code_file,
int id )

Initialize the CodedNetSDF object from model files.

This method initializes the CodedNetSDF object from two model files: one containing the trained neural network model and one containing the corresponding latent code.

Parameters
net_sdf_fileThe name of the file containing the trained neural network model.
latent_code_fileThe name of the file containing the corresponding latent code.
Examples
90_random_packing.cpp.

◆ PackJson()

nlohmann::json CodedNetSDF::PackJson ( )
overridevirtual

Pack the CodedNetSDF object into a JSON structure.

This method packs the CodedNetSDF object into a JSON structure for serialization and storage.

Returns
The packed JSON structure.

Reimplemented from netdem::Shape.

◆ Print()

void CodedNetSDF::Print ( )
overridevirtual

Print information about the CodedNetSDF object.

This method prints basic information about the CodedNetSDF object, including its position, rotation, size, and volume.

Reimplemented from netdem::Shape.

◆ SetSize()

void CodedNetSDF::SetSize ( double d)
overridevirtual

Set the size of the CodedNetSDF object.

This method sets the size of the CodedNetSDF object by scaling it uniformly in all dimensions.

Parameters
dThe new size of the CodedNetSDF object.

Reimplemented from netdem::Shape.

Examples
90_random_packing.cpp.

◆ SignedDistance()

double CodedNetSDF::SignedDistance ( Vec3d const & pos) const
overridevirtual

Compute the signed distance from a given position to the surface of the CodedNetSDF object.

This method computes the signed distance from a given position to the surface of the CodedNetSDF object by evaluating its neural network-generated signed distance function.

Parameters
posThe position for which to compute the signed distance.
Returns
The signed distance from the specified position to the surface of the CodedNetSDF object.

Reimplemented from netdem::Shape.

◆ SurfacePoint()

Vec3d CodedNetSDF::SurfacePoint ( Vec3d const & pos)
overridevirtual

Compute the surface point closest to a given position on the CodedNetSDF object.

This method computes the surface point closest to a given position on the CodedNetSDF object by using its neural network-generated signed distance function to perform an iterative search.

Parameters
posThe position for which to compute the closest surface point.
Returns
The surface point closest to the specified position on the CodedNetSDF object.

Reimplemented from netdem::Shape.

◆ UpdateShapeProperties()

void CodedNetSDF::UpdateShapeProperties ( )
overridevirtual

Update the shape properties of the CodedNetSDF object.

This method updates the shape properties of the CodedNetSDF object, such as its volume and surface area.

Reimplemented from netdem::Shape.

◆ UpdateSurfaceNodes()

void CodedNetSDF::UpdateSurfaceNodes ( )
overridevirtual

Update the surface nodes of the CodedNetSDF object.

This method updates the surface nodes of the CodedNetSDF object based on its current position and rotation.

Reimplemented from netdem::Shape.

Member Data Documentation

◆ latent_code

VecXT<double> netdem::CodedNetSDF::latent_code

The latent code used as input to the neural network model.

◆ net_sdf_model

RegressionNet netdem::CodedNetSDF::net_sdf_model

The neural network model used to generate the signed distance function.


The documentation for this class was generated from the following files: