hnuhyuwa / fullpointtransformer Goto Github PK

The fpconv code would be upgraded later.

• Ubuntu: 18.04 or higher

• PyTorch: 1.9.0

• CUDA: 11.1

• Hardware: 4GPUs (NVIDIA A40)

• To create conda environment, command as follows:

  sh env_setup.sh pt
  

• Download S3DIS dataset and symlink the paths to them as follows:

  mkdir -p dataset
  ln -s /path_to_s3dis_dataset dataset/s3dis
  

• Semantic segmantation on S3DIS Area 5

  • Train

    • Specify the gpu used in config and then do training:

      sh tool/train.sh s3dis fptransformer_repro
      

  • Test

    • Afer training, you can test the checkpoint as follows:

      CUDA_VISIBLE_DEVICES=0 sh tool/test.sh s3dis fptransformer_repro
      

  ---

## Acknowledgement
The code is based on [Point Transformer](https://arxiv.org/abs/2012.09164).