Skip to content

The second place solution to ModelNet-C classification in PointCloud-C Challenge 2022 (ECCV'22 Workshop).https://arxiv.org/abs/2210.15514

License

Notifications You must be signed in to change notification settings

zhulf0804/PV-Ada

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PV-Ada: Point-Voxel Adaptive Feature Abstraction for Robust Point Cloud Classification

This repository contains PyTorch implementation for PV-Ada: Point-Voxel Adaptive Feature Abstraction for Robust Point Cloud Classification.

News

Dataset

  • ModelNet40 [download]
    |- modelnet40_ply_hdf5_2048
        | - ply_data_train*.h5 (#5)
        | - ply_data_test*.h5 (#2)
        | - shape_names.txt
        | - train_files.txt
        | - test_files.txt
        | - ply_data_train*.json (#5)
        | - ply_data_test*.json (#2)
    
  • ModelNetC [download]
      |- modelnet_c
          | - clean.h5
          | - add_global*.h5 (#5)
          | - add_local*.h5 (#5)
          | - dropout_global*.h5 (#5)
          | - dropout_local*.h5 (#5)
          | - jitter*.h5 (#5)
          | - rotate*.h5 (#5)
          | - scale*.h5 (#5)
    
  • ExtraModelNetC [download]
    |- cls_extra_test_data.h5
    

Train

python train.py --pw --beta 1.0 --modelnet_root your_path_to_modelnet40 --modelnetc_root your_path_to_modelnetc --[tapering]

e.g. 
python train.py --pw --beta 1.0 --modelnet_root /mnt/ssd1/lifa_rdata/cls/modelnet40_ply_hdf5_2048 --modelnetc_root /mnt/ssd1/lifa_rdata/PointCloud-C/modelnet_c

python train.py --pw --beta 1.0 --modelnet_root /mnt/ssd1/lifa_rdata/cls/modelnet40_ply_hdf5_2048 --modelnetc_root /mnt/ssd1/lifa_rdata/PointCloud-C/modelnet_c --tapering

Evaluate

# For ModelNet40 test set
python evaluate.py --eval --ckpt your_path/model.t7 --modelnet_root your_path_to_modelnet40 

e.g.
python evaluate.py --eval --ckpt pretrained/modelnetc.t7 --modelnet_root /mnt/ssd1/lifa_rdata/cls/modelnet40_ply_hdf5_2048


# For ModelNet40-C Public test set
python evaluate.py --eval_corrupt --ckpt your_path/model.t7 --modelnetc_root your_path_to_modelnetc

e.g.
python evaluate.py --eval_corrupt --ckpt pretrained/modelnetc.t7 --modelnetc_root /mnt/ssd1/lifa_rdata/PointCloud-C/modelnet_c

Infer

python test.py --ckpt your_path/model.t7 --h5_path your_path_to_extra_modelnetc.h5 --saved_path your_path

e.g.
python test.py --ckpt pretrained/modelnetc.t7 --h5_path /mnt/ssd1/lifa_rdata/PointCloud-C/cls_extra_test_data.h5 --saved_path results/PCC 

Citation

If you find our work useful in your research, please consider citing:


Acknowledgements

ModelNet40-C, PointCloud-C, PCT and 3DeformRS.

About

The second place solution to ModelNet-C classification in PointCloud-C Challenge 2022 (ECCV'22 Workshop).https://arxiv.org/abs/2210.15514

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages