Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
PointTruss: K-Truss for Point Cloud Registration
Authors: Yue Wu, Jun Jiang, Yongzhe Yuan, Maoguo Gong, Qiguang Miao, Hao Li, Mingyang Zhang, wenping ma
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on KITTI, 3DMatch, and 3DLo Match demonstrate that our method consistently outperforms both traditional and learning-based approaches in various indoor and outdoor scenarios, achieving state-of-the-art results. |
| Researcher Affiliation | Academia | 1Mo E Key Lab of Collaborative Intelligence Systems, Xidian University 2School of Computer Science and Technology, Xidian University 3School of Electronic Engineering, Xidian University 4School of Artificial Intelligence, Xidian University 5Academy of AI, College of Mathematics Science, Inner Mongolia Normal University |
| Pseudocode | Yes | A.4 Pseudocode for Our Algorithm The following pseudocode outlines the complete pipeline of Point Truss, our robust 3D point cloud registration framework. |
| Open Source Code | No | We will make the complete code public following the acceptance of the paper. |
| Open Datasets | Yes | All datasets used in this work are publicly available. The Bunny model from the Stanford 3D Scanning Repository was acquired using a Cyberware 3030 MS scanner and is restricted to non-commercial use. The KITTI dataset is published under the Non Commercial-Share Alike 3.0 License and contains 11 sequences captured by a Velodyne HDL-64 3D Li DAR scanner in outdoor driving scenarios. [...] Additionally, we provide the 3DMatch dataset and its corresponding license information, as shown in Table 7, where 3DLo Match is a subset of 3DMatch. |
| Dataset Splits | Yes | For outdoor scenarios, we evaluate our method on the KITTI dataset [15]. Following the protocol established in [1, 5, 48], we select 555 point cloud pairs from sequences 8 to 10 for testing. For indoor environments, we conduct experiments on the 3DMatch dataset [45] and the more challenging 3DLo Match dataset [17], where point cloud pairs have less than 30% overlap. |
| Hardware Specification | Yes | All experiments are conducted on an AMD Ryzen 9 5950X CPU and a single NVIDIA RTX 3090 GPU. |
| Software Dependencies | No | Our method is implemented in Py Torch. |
| Experiment Setup | Yes | A.5 Hyper-parameter selection We set the inlier threshold τ to 0.1 for the 3DMatch and 3DLo Match datasets. For the KITTI dataset, τ is set to 0.6. The sampling ratio β ranges from 0.1 to 0.5. The k-truss parameter k is chosen between 3 and 10. The consensus threshold τc is set to 0.9 by default. The compatibility threshold τcomp is adjusted according to the noise standard deviation σ. The NMS radius rnms is typically set to 0.1. All hyperparameters are determined based on empirical validation. |