This repository contains the code of the paper "An Integer Linear Programming Approach to Geometrically Consistent Partial-Partial Shape Matching", V.Ehm, P. Roetzer, F. Bernard, D. Cremers. 3DV 2026. It provides code to solve the partial-to-partial shape matching problem with geometric consistency using an ILP.
Add a conda environment and install required packages
conda create -n partial-geco python=3.8
conda activate partial-geco
pip install -r requirements.txt
To compute surface cycles and generate the product space, we use the implementation of GeCo [1]. To use that run
pip install git+https://github.com/paul0noah/GeCo
Additionally install the Gurobi optimizer (www.gurobi.com) for solving the optimization problem.
Run the main.py to see an example shape being matched.
CP2P24 can be dowloaded from huggingface and PARTIALSMAL can be downladed from this GC-PPSM [2].
We use features and overlap predicions from EchoMatch [3]. You can find the repository here. Save the features and overlap predictions in example_data/features.
@inproceedings{ehminteger,
title={An Integer Linear Programming Approach to Geometrically Consistent Partial-Partial Shape Matching},
author={Ehm, Viktoria and Roetzer, Paul and Bernard, Florian and Cremers, Daniel},
booktitle={Thirteenth International Conference on 3D Vision}
}
[1] Paul Roetzer, Florian Bernard. Fast Globally Optimal and Geometrically Consistent 3D Shape Matching. ICCV, 2025.
[2] Viktoria Ehm, Maolin Gao, Paul Roetzer, Marvin Eisenberger, Daniel Cremers, Florian Bernard. Partial-to-Partial Shape Matching with Geometric Consistency. CVPR, 2024.
[3] Yizheng Xie, Viktoria Ehm, Paul Roetzer, Nafie El Amrani, Maolin Gao, Florian Bernard, Daniel Cremers. CVPR 2025.