Conquer3D v0.8.6
v0.8.6 · CUDA 12 · PyTorch 2.x

Conquer3D

A GPU-native toolbox for isosurface extraction, spatial acceleration, and volumetric conversion — written in CUDA, exposed as PyTorch tensors, and documented down to the individual kernel.

Get started API Reference GitHub
 dual_marching_cubes.py
import torch
from conquer3d.data_structure import create_voxel_grid
from conquer3d.ops import dmc

grid_vertices, voxels, _ = create_voxel_grid(
    grid_min=[-1.0] * 3, grid_max=[1.0] * 3,
    res=[64, 64, 64], device="cuda",
)

sdf = (torch.norm(grid_vertices, dim=-1) - 0.6).requires_grad_(True)
verts, faces = dmc(grid_vertices, voxels, sdf, iso=0.0)

verts.sum().backward()          # gradients flow into the field
Showcase

Real library output, by topic

6 figures

Mesh

Ray queries, curvature, quality metrics and orientation repair on triangle meshes.

View gallery →
9 figures

Isosurface

From a signed distance field to a surface: extraction, sharp features, resolution, normals and differentiable rendering.

View gallery →
4 figures

Others

Spatial ordering, nearest-neighbour search and learnable superquadric primitives.

View gallery →
Get started

Install and extract a surface

pip install -U conquer3d

Prebuilt CUDA wheels are published per release for Python 3.10–3.14 against PyTorch 2.8 and 2.11.

Read the guide → See the numbers →

import torch
from conquer3d.data_structure import create_voxel_grid
from conquer3d.ops import dmc

grid_vertices, voxels, _ = create_voxel_grid(
    grid_min=[-1.0] * 3, grid_max=[1.0] * 3,
    res=[64, 64, 64], device="cuda",
)

sdf = (torch.norm(grid_vertices, dim=-1) - 0.6).requires_grad_(True)
verts, faces = dmc(grid_vertices, voxels, sdf, iso=0.0)

verts.sum().backward()          # gradients flow into the field