RustedBytes

RustedBytes / manifold-knn

Public

Manifold k-NN: Accelerated k-NN Queries for Manifold Point Clouds

arxiv.orghtml2605.02224v1 geometry knn manifold nearest-neighbor point-cloud
10
0
89% credibility
Found Jun 01, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

manifold-knn is a Rust library that helps you find the closest points to any location in a large point cloud very quickly. It works by building a special spatial map that tracks how points are connected, then uses a clever search technique to explore only the most promising areas. The library supports adding and removing points dynamically, can work with 3D Delaunay triangulations, and offers optional speed boosts through parallel processing and SIMD instructions. It's based on a 2026 academic paper and provides mathematically exact results.

How It Works

1
📍 You have a collection of points

Your data exists as coordinates in space - like a 3D scan of an object, sensor readings, or spatial measurements.

2
🤔 You need to find nearby points quickly

Checking every single point to find the closest ones takes too long when you have thousands or millions of points.

3
⚡ Your index understands spatial connections

The library builds a special map that tracks how points are connected, so it can skip vast regions of space during searches.

4
Choose how to build your index
📊
Provide your own neighbor data

You already know which points were neighbors when your data was created.

🧊
Let the library build 3D connections

The library creates a Delaunay triangulation to discover how your points are spatially connected.

5
🔍 You search for the closest points

Ask 'which 5 points are nearest to this location?' and the index follows the most promising connections.

✅ You get fast, accurate results

The nearest neighbors appear in milliseconds, even with large datasets, and the results are mathematically exact.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is manifold-knn?

Manifold-knn is a Rust library for finding k-nearest neighbors in point clouds. It implements an academic paper's algorithm that uses dynamic programming to accelerate queries on manifold-structured data. Instead of checking every point, it follows a "successor table" built during insertion to prune the search space dramatically. The library ships with an optional 3D Delaunay triangulation backend for constructing these tables automatically, plus SIMD and parallel processing features for high-throughput workloads. You get exact k-NN results with support for prefix queries (searching a subset without rebuilding the index) and dynamic insertion and deletion.

Why is it gaining traction?

The killer feature is prefix queries: if your points arrive in order (temporal streams, coarse-to-fine scans), you can query any prefix of the data without rebuilding anything. For applications where data grows incrementally, this eliminates a common bottleneck. The built-in 3D Delaunay backend removes the need to wire up an external triangulation library, which is rare in this space. Rust's safety guarantees mean no unsafe code by default, which matters for geometry libraries where numerical edge cases bite you.

Who should use this?

Geometry developers working with 3D point clouds who need exact k-NN results, not approximate. Researchers benchmarking manifold learning algorithms will appreciate the exact-oracle fallback for validation. If you're building spatial indexes for CAD, LiDAR processing, or scientific visualization and need dynamic updates, this fits. It's not for general-purpose approximate nearest neighbor search; if you need that, look at HNSW libraries instead.

Verdict

This is a specialized tool for a specific problem class, and it looks solid within that scope. The documentation is thorough, the API is ergonomic, and the no-unsafe-code policy is refreshing. However, the repository has 10 stars and the credibility score is 0.8999999761581421%, indicating early-stage maturity. Test coverage appears adequate but the project is too young for production confidence. Worth evaluating now if your use case matches, but budget time for potential API changes before committing to a critical path.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.