Lightweight
Core source stays under 0.3 MB, ships under MIT, and drops cleanly into research or production trees.
Key capabilities at a glance
A compact, robust, and user-friendly lib for modern 3D registration tasks.
Efficient & robust
Fast-Robust-ICP and SPARE solvers help Lite3DRegLib outperform baseline ICP and CPD in speed and resilience.
Interactive exploration
The Gradio-powered interface on Hugging Face lets you tweak parameters, inspect residuals, and reset defaults live.
Accessible APIs
Unified Python and C++ bindings with curated datasets keep Windows and Linux experimentation straightforward.
Visualization interactively
Launch the Hugging Face Space to adjust registration parameters via dropdowns, explore example datasets, and reset to curated defaults while watching convergence unfold.
3D interaction tools
Spin, zoom, and compare rigid/non-rigid outputs in the same Gradio canvas.
Camera
Orbit & zoom
Hover
index & coordinates
Controls
Reset & save
Registration result
Rigid registration
Powered by Fast-Robust-ICP
Non-rigid registration
SPARE-based deformation
Build, integrate, and deploy fast
Dependencies, compilation routes, and the C++/Python interfaces that Lite3DRegLib provide.
Dependencies & build
- Eigen 3.4.0
- OpenMesh 8.1
- Optional (Linux): Intel MKL for accelerated solvers
pip install numpy gradio plyfile trimesh plotlycd scripts
./compile.sh # builds library + Lightweight3DRegDemo.exeC++ interface
RigidFricpRegistration* rigid;
NonrigidSpareRegistration* spare;
rigid->Reg(target_file, source_file, outpath);
spare->Reg(target_file, source_file, outpath);
rigid->Paras_init(use_init=false, file_init="", max_iter=80, stop=1e-5);
spare->Paras_init(30, 1e-3, 1e-4, false, {}, {});
rigid->Read_data(target_points, source_points, target_normals, source_normals);
spare->Read_data(target_points, source_points, target_normals, source_normals);
rigid->Register();
spare->Register();
rigid->Output_data(outpath, "rigid");
spare->Output_data(outpath, "spare");Python interface
import pyregister as l3dr
rigid = l3dr.RigidFricpRegistration()
spare = l3dr.NonrigidSpareRegistration()
rigid.Reg("data/target.ply", "data/source.ply", "./out/rigid/")
spare.Reg("data/target.obj", "data/source.obj", "./out/spare/")
rigid.Paras_init(useinit=False, fileinit="", maxiter=80, stop=1e-5)
spare.Paras_init(iters=30, stopcoarse=1e-3, stopfine=1e-4,
uselandmark=False, src=[], tar=[])
rigid.Paras_init(); spare.Paras_init()
rigid.Read_data(target_p, source_p, target_n, source_n)
spare.Read_data(target_p, source_p, target_n, source_n)
rigid.Register(); spare.Register()
rigid.Output_data("./out/rigid/", "rigid")
spare.Output_data("./out/spare/", "spare")Applications powered by Lite3DRegLib
Blend rigid and non-rigid pipelines to support robotics, cultural heritage, healthcare, and any workflow that relies on precise 3D alignment.
Autonomous fleets stabilize lidar and depth sweeps in real time, feeding SLAM loops with centimeter-level poses for navigation and mapping.
Digital twin and cultural heritage projects register terrestrial scans, photogrammetry meshes, and drone captures into unified asset vaults.
Medical imaging teams align anatomical scans and deformable surfaces to study motion, build patient-specific models, and track treatment progress.
Inside the Lite3DRegLib stack
The repository layout keeps core solvers, bindings, and tooling cleanly separated so you can adopt only the layers you need.
01
cpp/ core
Modern C++ solvers implementing Fast-Robust-ICP and SPARE algorithms with Eigen and templated utilities.
02
python/ bindings
High-level API mirrors the C++ entry points, enabling quick notebooks, scripting, and Hugging Face deployments.
03
examples/ & data
Sample datasets, reference pipelines, and walkthroughs for rigid and non-rigid registration experiments.
04
Gradio app.py
Ready-to-run visualization in Gradio with parameter controls, reset actions, and live residual monitoring.