This repository is a PyTorch implementation of Neural-Intrinsic-Embedding.
To install requirements:
pip install -r requirements.txt
Installing PyTorch may require an ad hoc procedure, depending on your computer settings.
You can find the data and the pre-trained models in:
data
models
To evaluate the model FAUST\SCAPE, run:
python code/faust/test_faust_sample.py
or
python code/scape/test_scape_sample.py
And in matlab the script:
code/eval/FAUST_5k.m
or
code/eval/SCAPE_5k.m
First, you should compute the geodesic distance matrix for each shape in the dataset. You can use the code in this repository or this matlab code and put them in:
data/{DATASET_NAME}/geod
For example:
data/SCAPE_5k/geod/mesh000.npy
if mesh000 has N points, then the distance matrix mesh000.npy has the shpae of [N,N].
To train the basis model, you may run:
python code/train_basis_sample.py --config config/train_scape_5k.yaml
Thenn, to train the descriptor model, you may run:
python code/train_desc_sample.py --config config/train_scape_5k.yaml
If you use this code, please cite our paper.
@inproceedings{jiang2023neural,
title={Neural Intrinsic Embedding for Non-rigid Point Cloud Matching},
author={Jiang, Puhua and Sun, Mingze and Huang, Ruqi},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={21835--21845},
year={2023}
}
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. For any commercial uses or derivatives, please contact us.