ModelNet

MoEx on ModelNet

This code is based on the example in PyTorch Geometric. We appreciate the PyTorch Geometric team for sharing their efficient implementation.

Requirements

This code was tested on the following versions, but they may not be necessary. Please follow the installation instructions in PyTorch Geometricl repo to install the packages.

torch>=1.3.0
torchvision>=0.4.2
torch-geometric>=1.3.2

Training a PointNet++ on ModelNet10 with MoEx (using instance normalization)

Note: there is a typo in our first draft (arXiv v1) saying that moex_lambda is set to 1 (i.e., not interpolating the labels), but in fact we use 0.9.

python pointnet2_classification_moex.py --moex_prob 0.5 --moex_lambda 0.9 --moex_norm "in" --data 10

Training a PointNet++ on ModelNet40 with MoEx (using instance normalization)

python pointnet2_classification_moex.py --moex_prob 0.5 --moex_lambda 0.9 --moex_norm "in" --data 40