This repository is for implementing frameworks introduced in the following paper:
"Backpropagation-free Network for 3D Test-time Adaptation"

Create a conda environment and install dependencies:

conda env create -f environment.yml

If this fails, please try to install the Point-NN and MATE environments one by one. 1.Point-NN 2.MATE

.
├── ...
├── datasets
├── data                    
│   ├── modelnet40_ply_hdf5_2048          
│   ├── modelnet40_c         
│   └── ...                
├── checkpoint                   
│   └── ... 
└── ....

Please download the following datasets for the prototype memory

Download ModelNet40

Download ScanObjectNN

Download ShapeNetCoreV2

Please follow the script from ModelNet40-C to add corruption.

We also provide download links:

Download ModelNet40-C from Google Drive.

Download ScanObjectNN-C from Google Drive.

checkpoint is uploaded here.

For ModelNet40 dataset, just run:

python run_BFTT3D_mn40.py --pth pointnet

The code is not fully organized and may not look polished. Refactoring and finalization are underway.