Runnable examples live in the examples/ directory. Run any with:
cargo run -p threecrate-examples --bin basic_usage
cargo run -p threecrate-examples --bin ransac_plane_example
# ...cargo run -p threecrate-examples --bin bevy_mesh_viewer --features bevy_interop
PBR-lit viewer built on Bevy. Press 1/2/3 to switch between Sphere, Torus, and Gyroid meshes (all generated live via Marching Cubes). Press C to cycle material colours, R to toggle auto-rotate. Drag to orbit, scroll to zoom.
cargo run -p threecrate-examples --bin bevy_conversion_demo --features bevy_interop
Converts a threecrate_core::TriangleMesh to a bevy::Mesh at runtime, prints the attribute details, then launches the Bevy viewer so you can inspect the result interactively.
cargo run -p threecrate-examples --bin interactive_viewer_example
Lightweight orbit/pan/zoom viewer using threecrate_visualization::InteractiveViewer — no Bevy dependency required.
cargo run -p threecrate-examples --bin gpu_mesh_render_example
wgpu-based real-time mesh renderer.
cargo run -p threecrate-examples --bin basic_usageThreeCrate Basic Usage Example
=============================
1. Basic ICP Registration
-------------------------
Source points : 1000
Target points : 1000
ICP converged in 23 iterations
Final RMSE : 0.001247
2. ICP with Known Transformation
--------------------------------
Applied rotation: 45° around Z, translation (1, 0, 0)
ICP converged in 18 iterations
Rotation error : 0.0003 rad
Translation error: 0.0008 m
3. ICP with Noise
-----------------
Noise σ=0.01 — converged in 31 iterations, RMSE: 0.0089
cargo run -p threecrate-examples --bin ransac_plane_example=== RANSAC Plane Segmentation Example ===
1. Simple planar point cloud:
Input points: 10000
Found inliers: 9743
Plane coefficients: [0.0000, 0.0000, 1.0000, 0.0000]
Inlier percentage: 97.4%
2. Noisy planar point cloud:
Input points: 10000
Found inliers: 7821
Inlier percentage: 78.2%
cargo run -p threecrate-examples --bin euclidean_cluster_example=== Euclidean Cluster Extraction Example ===
1. Three well-separated blobs:
Input points : 3000
Clusters found: 3
Cluster 1: 1000 points
Cluster 2: 1000 points
Cluster 3: 1000 points
2. Parallel extraction on a denser scene:
Input points : 50000
Clusters found: 5
cargo run -p threecrate-examples --bin mesh_smoothing_example=== Mesh Smoothing Example ===
Input mesh: 25 vertices, 32 faces
Spike vertex z before smoothing: 2.0000
1. Laplacian (λ=0.5, 10 iters):
Spike vertex z: 0.3821
Spread: 0.0412
2. Taubin (λ=0.5, μ=-0.53, 10 iters):
Spike vertex z: 0.1204
Spread: 0.0387
3. HC smoothing (α=0.5, β=0.5, 10 iters):
Spike vertex z: 0.0973
Spread: 0.0401
cargo run -p threecrate-examples --bin normal_estimation_exampleEnhanced Normal Estimation Example
===================================
✓ Created sample point cloud with 5000 points
1. Basic k-NN Normal Estimation
--------------------------------
✓ Estimated normals using k-NN in 12.3ms
- Result: 5000 points with normals
2. Radius-based Normal Estimation
-----------------------------------
✓ Estimated normals using radius search in 18.7ms
- Result: 5000 points with normals
cargo run -p threecrate-examples --bin k_nearest_neighbors_example=== K-Nearest Neighbors Example ===
Created point cloud with 1000 points
Added 20 random points. Total: 1020 points
1. Finding k-nearest neighbors for each point:
Found 5 nearest neighbors for each point:
Point 0: [1, 4, 7, 2, 9]
Point 1: [0, 4, 2, 7, 3]
... and 1018 more points
cargo run -p threecrate-examples --bin simd_distance_example=== SIMD Distance Computation Example ===
--- Batch distance computation (16 points) ---
Max SIMD vs scalar error: 0.00e0 (should be ~0)
First 4 squared distances: [1.23, 4.56, 0.89, 2.34]
--- k-NN search comparison ---
SIMD k=3 nearest: [2, 5, 8]
Scalar k=3 nearest: [2, 5, 8]
Results match: true
cargo run -p threecrate-examples --bin mesh_boolean_example=== Mesh Boolean Operations Example ===
Cube A: 8 verts, 12 faces
Cube B: 8 verts, 12 faces
(The two cubes overlap in the region [1,2]^3)
1. Union (A ∪ B):
20 verts, 36 faces
2. Intersection (A ∩ B):
8 verts, 12 faces
3. Difference (A − B):
14 verts, 24 faces
cargo run -p threecrate-examples --bin test_marching_cubesTesting Complete Marching Cubes Implementation
==============================================
Test 1: Extracting isosurface from a sphere...
✓ Successfully extracted mesh from sphere
- Vertices: 1284
- Faces: 2564
- Has normals: true
Test 2: Extracting isosurface from a torus...
✓ Successfully extracted mesh from torus
- Vertices: 3812
- Faces: 7620
cargo run -p threecrate-examples --bin streaming_pipeline=== Streaming Point Cloud Pipeline Example ===
Synthetic dataset: 500000 points
In-memory size: ~6.0 MB
--- Pass 1: Streaming statistics (chunk_size=10 000) ---
Points processed : 500000
Bounding box : (-5.00,-5.00,-5.00) – (5.00,5.00,5.00)
Mean position : (0.00,0.01,-0.01)
Chunks processed : 50 (43.21 ms total)
Throughput : 11.6 Mpts/s
--- Pass 2: Streaming voxel filter (voxel_size=0.05) ---
Output points : 38412
Reduction ratio : 92.3%
Elapsed : 81.4 ms
cargo run -p threecrate-examples --bin gpu_example
cargo run -p threecrate-examples --bin gpu_filtering_exampleThese examples run GPU-accelerated compute passes (voxel filtering, nearest-neighbor) and print timing results to the terminal. No window is opened.
import numpy as np
import threecrate as tc
cloud = tc.read_point_cloud("scan.ply")
cloud = tc.voxel_downsample(cloud, voxel_size=0.02)
cloud = tc.remove_statistical_outliers(cloud)
# Segment dominant plane
result = tc.segment_plane(cloud, threshold=0.01)
plane_cloud = result.inlier_cloud(cloud)
# Cluster remaining points
clusters = tc.extract_clusters(cloud)
print(f"{len(clusters)} objects found")
# Reconstruct surface
normal_cloud = tc.estimate_normals(cloud)
mesh = tc.poisson_reconstruct(normal_cloud)
mesh = tc.smooth_mesh_taubin(mesh, iterations=10)
mesh = tc.simplify_mesh(mesh, reduction_ratio=0.5)
tc.write_mesh(mesh, "output.ply")For the full Python API see threecrate-python/README.md.