Template3D-AD: Point Cloud Template Matching Method Based on Center Points for 3D Anomaly Detection

April 30, 2025 · View on GitHub

Yi Liu*, Changsheng Zhang*, Yufei Yang

(* Corresponding authors)

Our paper has been accepted by IJCAI 2025

Overview

This paper proposes Template3D-AD, the first 3D anomaly detection method based on template matching.

Data preparation

run save_fpfh.py

voxel_size = 0.25 for Real3D-AD dataset and voxel_size = 0.25 for Anomaly-ShapeNet-v2 dataset

Run

run main.py

Acknowledgments.

This work is supported by the Fundamental Research Funds for the Central Universities (N25BSS022) and the Ningxia Natural Science Foundation (2024AAC03349).

Our benchmark is built on M3DM and Real3D-AD, thanks their extraordinary works!

BibTex Citation

If you find this paper and repository useful, please cite our paper☺️.