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

#include <Model.h>

Public Member Functions
	Model ()
	Model (float voxel_size, int block_resolution, int block_count, 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 ()
void	UpdateFramePose (int frame_id, const core::Tensor &T_frame_to_world)
void	SynthesizeModelFrame (Frame &raycast_frame, float depth_scale, float depth_min, float depth_max, bool enable_color=true)
odometry::OdometryResult	TrackFrameToModel (const Frame &input_frame, const Frame &raycast_frame, float depth_scale, float depth_max, float depth_diff)
void	Integrate (const Frame &input_frame, float depth_scale, float depth_max)
t::geometry::PointCloud	ExtractPointCloud (int estimated_number=-1, float weight_threshold=3.0f)
t::geometry::TriangleMesh	ExtractTriangleMesh (int estimated_number=-1, float weight_threshold=3.0f)
core::HashMap	GetHashMap ()
	Get block hashmap int the TSDFVoxelGrid. More...

Data Fields
t::geometry::VoxelBlockGrid	voxel_grid_
	Maintained volumetric map. More...
core::Tensor	frustum_block_coords_
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,
		int	block_count,
		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	(	int	estimated_number = -1,
		float	weight_threshold = 3.0f
	)

Extract surface point cloud for visualization / model saving.

Parameters

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

Returns

Extracted point cloud.

ExtractTriangleMesh()

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

Extract surface triangle mesh for visualization / model saving.

Parameters

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

Returns

Extracted point cloud.

GetCurrentFramePose()

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

inline

GetHashMap()

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

(

)

Get block hashmap int the TSDFVoxelGrid.

Integrate()

void open3d::t::pipelines::slam::Model::Integrate	(	const Frame &	input_frame,
		float	depth_scale,
		float	depth_max
	)

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.

SynthesizeModelFrame()

void open3d::t::pipelines::slam::Model::SynthesizeModelFrame	(	Frame &	raycast_frame,
		float	depth_scale,
		float	depth_min,
		float	depth_max,
		bool	enable_color = true
	)

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.

TrackFrameToModel()

odometry::OdometryResult open3d::t::pipelines::slam::Model::TrackFrameToModel	(	const Frame &	input_frame,
		const Frame &	raycast_frame,
		float	depth_scale,
		float	depth_max,
		float	depth_diff
	)

Track using PointToPlane depth odometry. TODO(wei): support hybrid or intensity methods.

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.

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_

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 (4dea5be (Tue Nov 30 18:43:52 2021 -0800))
• /home/runner/work/Open3D/Open3D/cpp/open3d/t/pipelines/slam/Model.cpp (4dea5be (Tue Nov 30 18:43:52 2021 -0800))