Point-to-Point Correspondence

This repository facilitates surface point-to-point correspondence utilizing the ZoomOut method. It wraps the ZoomOut implementation from PyFM and exposes the get_correspondence function for making predictions. Configuration options are provided through the get_default_config function, offering a dictionary of default parameters that can be passed to get_correspondence.

Installation

You can install the package using pip:

pip install git+github.com/shaharzuler/flow_n_corr_utils.git

Usage

Here's a basic usage example:

import p2p_correspondence
from easydict import EasyDict

mesh1_path = "path/to/mesh1.off"
mesh2_path = "path/to/mesh2.off"
# Generate default configuration, read config file from src/default_config.json
config = p2p_correspondence.get_default_config()
# Perform point correspondence inference
inference_path, flow_is_valid = p2p_correspondence.get_correspondence(mesh1_path, mesh2_path, EasyDict(config))

For a detailed usage example, please refer to run.py.

Outputs

The model inference generates the following outputs:

• P (Point Correspondence Confidence Matrix): Both point clouds along with the confidence matrix are written to an HDF5 file. You can use the get_point_clouds_and_p_from_h5 function to read the file.
• Visualization: Corresponding .obj files for both point clouds with color representations are saved in the output folder. Additionally, the confidence matrix C visualization is saved.

Example for C matrix:

The confidence matrix visualizes the reliability of correspondences between points represented by functional maps (FM) on both surfaces, with values ranging from 0 (dark, low confidence) to 1 (yellow, high confidence). I

Example for Color Representation for Correspondence:

Each point on the first point cloud is colored the same as its corresponding point on the second point cloud.

Full Implementation

For a comprehensive implementation of our paper, utilizing this package along with additional functionalities, please refer to CardioSpectrum implementation.