Point Cloud Completion via Skeleton-Detail Transformer

Codes for Point Cloud Completion via Skeleton-Detail Transformer. IEEE Transactions on Visualization and Computer Graphics (TVCG), 2022. See IEEE PDF.

In this work, we present a coarse-to-fine completion framework, which makes full use of both neighboring and long-distance region cues for point cloud completion. Our network leverages a Skeleton-Detail Transformer, which contains cross-attention and self-attention layers, to fully explore the correlation from local patterns to global shape and utilize it to enhance the overall skeleton. Also, we propose a selective attention mechanism to save memory usage in the attention process without significantly affecting performance.

1) Pre-requisites

• Python3
• CUDA
• pytorch
• open3d-python

This code is built using Pytorch 1.7.1 with CUDA 10.2 and tested on Ubuntu 18.04 with Python 3.6.

2)Compile 3rd-party libs

The libs are included under /util, you need to first compile them where there is also a 'Readme.md' in each subfolder.

3)Download pre-trained models

Download pre-trained models in trained_model folder from Google Drive and put them on trianed_model dir.

4) Testing

For PCN:

1. Download ShapeNet test data on Google Drive. Put them on data/pcn folder. We use the same testing data in PCN project but we use h5 format.
2. Run sh test.sh. You should first modify the model_path to the folder containing your pre-trained model, and data_path to the testing files.

For Completion3D:

1. Download the test data on Google Drive or Completion3D. Put them on data/completion3d folder.
2. Run test_benchmark.sh to generate the 'submission.zip' file for Compleiont3D dataset.

5) Traning

For PCN

1. The training data are from PCN repository, you can download training (train.lmdb, train.lmdb-lock) and validation (valid.lmdb, valid.lmdb-lock) data from shapenet directory on the provided training set link in PCN repository.
2. Run python create_pcn_h5.py to generate the training and validation files with .h5 format.
3. Run sh run.sh for training.

For Compleiont3D: You can directly download the tranining files from Compleiont3D benchmark. Run sh run.sh and set dataset to Completion3D.

[Acknowledgement]

Our codes are partly from ECG, VRCNET. We sincerely thank for their contribution.