Semantics-Augmented Quantization-Aware Training for Point Cloud Classification

Introduction

This project is the official implementation of our paper Semantics-Augmented Quantization-Aware Training for Point Cloud Classification.

Installation

The latest codes are tested on Ubuntu 22.04, CUDA11.7, PyTorch 1.13.1 and Python 3.8:

# create new conda environment
conda create -n saqat python=3.8 -y
conda activate saqat

# install pytorch
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia

# install pytorch3d
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install pytorch3d -c pytorch3d

# install other dependencies
pip install brevitas
pip install h5py
pip install opencv-python

Dataset

ModelNet40

Note in this experiment, we do not use any re-sampled version of ModelNet40 (more than 2K points) or any normal information. The dataset utilized here is: modelnet40_ply_hdf5_2048. ModelNet40 dataset will be downloaded automatically.

ScanObjectNN

The hardest variant is available for download via the provided link. Please cite their paper if you use the link to download the data.

mkdir data
cd data
gdown https://drive.google.com/uc?id=1iM3mhMJ_N0x5pytcP831l3ZFwbLmbwzi
tar -xvf ScanObjectNN.tar

Organize the ScanObjectNN dataset as follows:

data
 |--- h5_files
        |--- main_split
                |--- training_objectdataset_augmentedrot_scale75.h5
                |--- test_objectdataset_augmentedrot_scale75.h5

Train

conda activate saqat
python train_classification.py --gpu 0 --model pointnet_quant_cls --dataset scanobjectnn --log_dir SONN_cls --quant_bit 8 --knn_k 30 --max_probability 0.1 --loss2_weight 1 --learning_rate 0.0005

Test

conda activate saqat
python test_classification.py --gpu 1 --model pointnet_quant_cls --dataset scanobjectnn --log_dir SONN_cls