PSVMLP: point and shifted voxel MLP For 3D deep learning

by Guanghu Xie, Yang Liu*, Yiming Ji, Zongwu Xie, Baoshi Cao

Introduction

This repository is built for the official implementation of:

PAConv: point and shifted voxel MLP For 3D deep learning

Object Classification(ModelNet40)

Dataset

Download and unzip ModelNet40 (415M). Then symlink the paths to it as follows (you can alternatively modify the path here):

mkdir -p data
ln -s /path to modelnet40/modelnet40_ply_hdf5_2048 data

Train:

python main.py --config config/psvmlp_train.yaml

Test:

python eval_voting.py --config config/psvmlp_test.yaml`

Object Classification(ScanObjectNN)

The dataset will be automatically downloaded from ScanObjectNN

cd classification_ScanObjectNN
# train
python main_part_seg.py --model psvmlp

By default, it will create a fold named "checkpoints/{modelName}-{msg}-{randomseed}", which includes args.txt, best_checkpoint.pth, last_checkpoint.pth, log.txt, out.txt.

Part segmentation(ShapeNet)

• Make data folder and download the dataset from ShapeNet.

cd part_segmentation
mkdir data
cd data
wget https://shapenet.cs.stanford.edu/media/shapenetcore_partanno_segmentation_benchmark_v0_normal.zip --no-check-certificate
unzip shapenetcore_partanno_segmentation_benchmark_v0_normal.zip

• Train PSVMLP

# train PSVMLP
python main.py --config config/psvmlp_partseg_train.yaml
# please add other paramemters as you wish.

• Test PSVMLP

# test PSVMLP
python main.py --config config/psvmlp_partseg_test.yaml
# please add other paramemters as you wish.

Acknowledgment

Our implementation is mainly based on the following codebases. We gratefully thank the authors for their wonderful works.

PointMLP, PAConv,

LICENSE

PSVMLP is under the Apache-2.0 license. # psvmlp