Quick Start

Extract a surface in a few lines of code.

import torch
import isoext
from isoext.sdf import SphereSDF, TorusSDF, UnionOp, IntersectionOp, NegationOp, RotationOp
from isoext import viewer

# A high-resolution grid of sample points
grid = isoext.UniformGrid([256, 256, 256])

# A shape built with CSG: a sphere with three interlocking toroidal tunnels
sphere = SphereSDF(radius=0.75)
torus = TorusSDF(R=0.75, r=0.15)
shape = IntersectionOp(
    [
        sphere,
        NegationOp(
            UnionOp(
                [
                    torus,
                    RotationOp(torus, axis=[1, 0, 0], angle=90),
                    RotationOp(torus, axis=[0, 1, 0], angle=90),
                ]
            )
        ),
    ]
)

# Evaluate the shape at every grid point and extract the surface
grid.set_values(shape(grid.get_points()))
vertices, faces = isoext.marching_cubes(grid)

print(f"Extracted {vertices.shape[0]:,} vertices, {faces.shape[0]:,} triangles")
viewer.embed(vertices, faces, color="coral")

Extracted 196,176 vertices, 392,348 triangles

What Just Happened

The example built a 256³ grid of sample points, evaluated a signed distance function at each point, and ran marching cubes to mesh the surface where the field crosses zero. If any of those words are new, What Is Iso-Surface Extraction? explains them from scratch; Method Comparisons compares the methods.

Interactive Viewing

Outside of these docs, isoext.viewer can also open a live viewer in your browser:

server = viewer.show(vertices, faces)  # prints the URL it serves on
# ... inspect the mesh in the browser ...
server.stop()

viewer.embed, used above, instead records the scene to a static file next to the notebook, so the rendered page stays interactive without a running server.

Learn More