Visualization
reconstruct_surface_marching_cubes(model, domain, device, iso=0, res=256, batch_size=2 ** 16, **kwargs)
Extracts isosurfaces of a given neural implicit by using the Marching Cube algorithm.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model
|
Module
|
the neural implicit function. |
required |
domain
|
AABB
|
domain onto which the function values are computed. |
required |
device
|
device
|
device onto which the computation are performed. |
required |
iso
|
int
|
value of the isosurface. Several values can be provided in a list. If that is the case, one mesh per isovalue will be produced. Defaults to 0. |
0
|
res
|
int
|
resolution of the marching cubes grid. Defaults to 100. |
256
|
batch_size
|
int
|
batch size for forward computation. Defaults to 2**16. |
2 ** 16
|
Other Parameters:
| Name | Type | Description |
|---|---|---|
use_tqdm |
bool
|
whether to display a progress bar during computation. Defaults to False. |
full_precision |
bool
|
by default, the computation is done over torch.float16 for faster inference. This flag enables computation in full torch.float32 precision. Defaults to False. |
Returns:
| Name | Type | Description |
|---|---|---|
dict |
dict
|
returns a dictionary of ((i,iso), mesh) where i is a unique identifier, iso is the value of the isovalue and mesh is the output surface mesh (mouette.mesh.SurfaceMesh type). |
References
- Marching cubes: A high resolution 3D surface construction algorithm, Lorensen & Cline (1998)
- https://scikit-image.org/docs/0.25.x/auto_examples/edges/plot_marching_cubes.html
render_lip_constant_2d(img_path, model, domain, device, res=1000, batch_size=2 * 15)
Renders a 2D image of the local Lipschitz constant (maximal gradient norm in a neighborhood) of a 2D model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
img_path
|
str
|
path to export the render. |
required |
model
|
Module
|
the neural model |
required |
domain
|
AABB
|
domain onto which the function values are computed. |
required |
device
|
device
|
device onto which the computation are performed. |
required |
res
|
int
|
Image resolution. Defaults to 1000. |
1000
|
batch_size
|
int
|
batch size for forward computation. Defaults to 2**15. |
2 * 15
|
render_sdf_2d(render_path, contour_path, gradient_path, model, domain, device, res=1000, batch_size=2 ** 15, **kwargs)
Renders a 2D SDF
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
render_path
|
str
|
Path to export a color render |
required |
contour_path
|
str
|
Path to export the contour plot |
required |
gradient_path
|
str
|
Path to export the gradient norm plot |
required |
model
|
SDF neural model |
required | |
domain
|
AABB
|
axis aligned bounding box of dimension 2to represent the plotting domain. |
required |
device
|
device
|
device onto which the computation are performed. |
required |
res
|
int
|
Image resolution. Defaults to 800. |
1000
|
batch_size
|
int
|
Size of forward batches. Defaults to 2**15. |
2 ** 15
|
Other Parameters:
| Name | Type | Description |
|---|---|---|
n_contours |
int
|
Defaults to 16. |
render_sdf_quad(render_path, contour_path, gradient_path, model, P0, P1, P2, device, res=1000, batch_size=2 ** 15, **kwargs)
Renders a 3D sdf along a specified (planar) quad in space.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
render_path
|
str
|
Path to export a color render |
required |
contour_path
|
str
|
Path to export the contour plot |
required |
gradient_path
|
str
|
Path to export the gradient norm plot |
required |
model
|
SDF neural model |
required | |
P0
|
Tensor
|
First point of the quad ( (0,0) in parameter space) |
required |
P1
|
Tensor
|
Second point of the quad ( (1,0) in parameter space) |
required |
P2
|
Tensor
|
Third point of the quad ( (0,1) in parameter space) |
required |
device
|
device
|
device onto which the computation are performed. |
required |
res
|
int
|
Image resolution. Defaults to 1000. |
1000
|
batch_size
|
int
|
Size of forward batches. Defaults to 2**15. |
2 ** 15
|