Skip to content

ZhaofanQiu/SPE-Net

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SPE-Net: Boosting Point Cloud Analysis via Rotation Robustness Enhancement

This repository is for SPE-Net: Boosting Point Cloud Analysis via Rotation Robustness Enhancement. The repository is based on the open-source codebase Detectron2 and CloserLook3D.

Requiremenets

  • Linux with Python ≥ 3.6
  • PyTorch ≥ 1.8
  • fvcore
  • pycocotools
  • Java 1.8.0

Installation

  • Prepare the target dataset following the instructions from the codebase CloserLook3D, and put the pre-processed data in the folder ./dataset.
  • Run pytorchpoints/init.sh to compile the C++ code.

Train SPE-Net Model

configs/spe_net_modelnet40_so3.yaml is the config file for training SPE-Net model on ModelNet40. Run script python3 train_net.py --num-gpus 4 --config-file configs/spe_net_modelnet40_so3.yaml.

pytorchpoints/modeling/backbones/resnet.py includes the implementation for SPE-Net overall architecture.

pytorchpoints/modeling/local_aggregation/spe_mlp.py includes the implementation for SPE-MLP.

Acknowledgements

Thanks the contribution of Detectron2, CloserLook3D and the PyTorch team.

About

Code release for ECCV22 paper "SPE-Net: Boosting Point Cloud Analysis via Rotation Robustness Enhancement"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published