DDM: A Metric for Comparing 3D Shapes Using Directional Distance Fields (T-PAMI 2025)

[Arxiv] [T-PAMI]

Installation

cd DDM
python set_up.py install

Usage

Here we provide three versions of DDM, i.e., Point-to-Point, Mesh-to-Mesh, and Mesh-to-Point (in the paper, it is face).

Optimize a point cloud according to the reference point cloud

from Closest_Point_on_Surface import DDM_P2P

loss_func=DDM_P2P()

......
loss=loss_func(src_points,tgt_points)
......

Optimize a triangle mesh according to the reference triangle mesh

from Closest_Point_on_Surface import DDM_M2M

loss_func=DDM_M2M()

......
loss=loss_func(src_v,src_f,tgt_v,tgt_f)
......

Optimize a triangle mesh according to the reference point cloud

from Closest_Point_on_Surface import DDM_M2P

loss_func=DDM_M2P()

......
loss=loss_func(src_v,src_f,tgt_points)
......

TODO

• Release code.
• Release examples on the selected applications
  • Shape Deformation
  • Rigid Registration
  • Non-rigid Registration
  • Scene Flow Estimation
  • Human Pose Optimization

Acknowledgement

We thank the following excellent works including MAC, AMM, NSFP, SCOOP, and MDA for their released codes.

Citation

@article{ren2024measuring,
  title={DDM:  A Metric for Comparing 3D Shapes Using Directional Distance Fields},
  author={Ren, Siyu and Hou, Junhui and Chen, Xiaodong and Xiong, Hongkai and Wang, Wenping},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2025}
}