open3d::t::pipelines::slam::Model Class Reference

#include <Model.h>

Public Member Functions
	Model ()
	Model (float voxel_size, int block_resolution=16, int block_count=1000, const core::Tensor &T_init=core::Tensor::Eye(4, core::Float64, core::Device("CPU:0")), const core::Device &device=core::Device("CUDA:0"))
core::Tensor	GetCurrentFramePose () const
void	UpdateFramePose (int frame_id, const core::Tensor &T_frame_to_world)
void	SynthesizeModelFrame (Frame &raycast_frame, float depth_scale=1000.0, float depth_min=0.1, float depth_max=3.0, float trunc_voxel_multiplier=8.0, bool enable_color=true, float weight_threshold=-1.0)
odometry::OdometryResult	TrackFrameToModel (const Frame &input_frame, const Frame &raycast_frame, float depth_scale=1000.0, float depth_max=3.0, float depth_diff=0.07, odometry::Method method=odometry::Method::PointToPlane, const std::vector< odometry::OdometryConvergenceCriteria > &criteria={6, 3, 1})
void	Integrate (const Frame &input_frame, float depth_scale=1000.0, float depth_max=3.0, float trunc_voxel_multiplier=8.0f)
t::geometry::PointCloud	ExtractPointCloud (float weight_threshold=3.0f, int estimated_number=-1)
t::geometry::TriangleMesh	ExtractTriangleMesh (float weight_threshold=3.0f, int estimated_number=-1)
core::HashMap	GetHashMap ()
	Get block hashmap int the VoxelBlockGrid. More...

Data Fields
t::geometry::VoxelBlockGrid	voxel_grid_
	Maintained volumetric map. More...
core::Tensor	frustum_block_coords_
	Active block coordinates from prior integration. More...
core::Tensor	T_frame_to_world_
int	frame_id_ = -1

Detailed Description

Model class maintaining a volumetric grid and the current active frame's pose. Wraps functionalities including integration, ray casting, and surface reconstruction.

Constructor & Destructor Documentation

Model() [1/2]

open3d::t::pipelines::slam::Model::Model ( )

inline

Model() [2/2]

open3d::t::pipelines::slam::Model::Model	(	float	voxel_size,
		int	block_resolution = 16,
		int	block_count = 1000,
		const core::Tensor &	T_init = core::Tensor::Eye(4,                                                         core::Float64,                                                         core::Device("CPU:0")),
		const core::Device &	device = core::Device("CUDA:0")
	)

Member Function Documentation

ExtractPointCloud()

t::geometry::PointCloud open3d::t::pipelines::slam::Model::ExtractPointCloud	(	float	weight_threshold = 3.0f,
		int	estimated_number = -1
	)

Extract surface point cloud for visualization / model saving.

Parameters

weight_threshold	Weight threshold of the TSDF voxels to prune noise.
estimated_number	Estimation of the point cloud size, helpful for real-time visualization.

Returns

Extracted point cloud.

ExtractTriangleMesh()

t::geometry::TriangleMesh open3d::t::pipelines::slam::Model::ExtractTriangleMesh	(	float	weight_threshold = 3.0f,
		int	estimated_number = -1
	)

Extract surface triangle mesh for visualization / model saving.

Parameters

weight_threshold	Weight threshold of the TSDF voxels to prune noise.
estimated_number	Estimation of the point cloud size, helpful for real-time visualization.

Returns

Extracted point cloud.

GetCurrentFramePose()

core::Tensor open3d::t::pipelines::slam::Model::GetCurrentFramePose ( ) const

inline

GetHashMap()

core::HashMap open3d::t::pipelines::slam::Model::GetHashMap

(

)

Get block hashmap int the VoxelBlockGrid.

Integrate()

void open3d::t::pipelines::slam::Model::Integrate	(	const Frame &	input_frame,
		float	depth_scale = 1000.0,
		float	depth_max = 3.0,
		float	trunc_voxel_multiplier = 8.0f
	)

Integrate RGBD frame into the volumetric voxel grid.

Parameters

input_frame	Input RGBD frame.
depth_scale	Scale factor to convert raw data into meter metric.
depth_max	Depth truncation to discard points far away from the camera.
trunc_voxel_multiplier	Truncation distance multiplier in voxel size for signed distance. For instance, –trunc_voxel_multiplier=8 with –voxel_size=0.006(m) creates a truncation distance of 0.048(m).

SynthesizeModelFrame()

void open3d::t::pipelines::slam::Model::SynthesizeModelFrame	(	Frame &	raycast_frame,
		float	depth_scale = 1000.0,
		float	depth_min = 0.1,
		float	depth_max = 3.0,
		float	trunc_voxel_multiplier = 8.0,
		bool	enable_color = true,
		float	weight_threshold = -1.0
	)

Apply ray casting to obtain a synthesized model frame at the down sampled resolution.

Parameters

raycast_frame	RGBD frame to fill the raycasting results.
depth_scale	Scale factor to convert raw data into meter metric.
depth_min	Depth where ray casting starts from.
depth_max	Depth where ray casting stops at.
trunc_voxel_multiplier	Truncation distance multiplier in voxel size for signed distance. For instance, –trunc_voxel_multiplier=8 with –voxel_size=0.006(m) creates a truncation distance of 0.048(m).
enable_color	Enable color in the raycasting results.
weight_threshold	Weight threshold of the TSDF voxels to prune noise. Use -1 to apply default logic: min(frame_id_ * 1.0f, 3.0f).

TrackFrameToModel()

odometry::OdometryResult open3d::t::pipelines::slam::Model::TrackFrameToModel	(	const Frame &	input_frame,
		const Frame &	raycast_frame,
		float	depth_scale = 1000.0,
		float	depth_max = 3.0,
		float	depth_diff = 0.07,
		odometry::Method	method = odometry::Method::PointToPlane,
		const std::vector< odometry::OdometryConvergenceCriteria > &	criteria = {6, 3, 1}
	)

Track using depth odometry.

Parameters

input_frame	Input RGBD frame.
raycast_frame	RGBD frame generated by raycasting.
depth_scale	Scale factor to convert raw data into meter metric.
depth_max	Depth truncation to discard points far away from the camera
method	Method used to apply RGBD odometry.
criteria	Criteria used to define and terminate iterations. In multiscale odometry the order is from coarse to fine. Inputting a vector of iterations by default triggers the implicit conversion.

UpdateFramePose()

void open3d::t::pipelines::slam::Model::UpdateFramePose ( int  frame_id, const core::Tensor &  T_frame_to_world  )

inline

Field Documentation

frame_id_

int open3d::t::pipelines::slam::Model::frame_id_ = -1

frustum_block_coords_

core::Tensor open3d::t::pipelines::slam::Model::frustum_block_coords_

Active block coordinates from prior integration.

T_frame_to_world_

core::Tensor open3d::t::pipelines::slam::Model::T_frame_to_world_

T_frame_to_model, maintained tracking state in a (4, 4), Float64 Tensor on CPU.

voxel_grid_

t::geometry::VoxelBlockGrid open3d::t::pipelines::slam::Model::voxel_grid_

Maintained volumetric map.

---

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

• /home/runner/work/Open3D/Open3D/cpp/open3d/t/pipelines/slam/Model.h (1e7b174 (Tue Jan 7 15:20:18 2025 -0800))
• /home/runner/work/Open3D/Open3D/cpp/open3d/t/pipelines/slam/Model.cpp (1e7b174 (Tue Jan 7 15:20:18 2025 -0800))