Standalone

KPConvX: Standalone code

This folder contains a standalone version of KPConvX code.

Setup

Our code was tested with multiple environments and should be straightforward to setup. Addapt the following lines to your environment and version of CUDA:

conda create -n kpconvx python=3.10
conda activate kpconvx
conda install pytorch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 pytorch-cuda=12.1 -c pytorch -c nvidia
pip install easydict h5py matplotlib numpy scikit-learn timm pykeops
pip install 'pyvista[all,trame]' jupyterlab

Prepare data

S3DIS

We use the preprocessed data from Pointcept that you can download here. Please agree with the official license before downloading it. From the s3dis.tar.gz archive, extract the s3dis folder to the Standalone/data directory.

ScanObjectNN

We use the h5_files from official website. Download the preprocessed version and place it in the Standalone/data directory. Rename the folder h5_files to ScanObjectNN. You should have (and only need) the three following files:

data/ScanObjectNN/main_split/test_objectdataset_augmentedrot_scale75_1024_fps.pkl
data/ScanObjectNN/main_split/test_objectdataset_augmentedrot_scale75.h5
data/ScanObjectNN/main_split/training_objectdataset_augmentedrot_scale75.h5

Experiments

Training networks

We provide scripts to train our model on ScanObjectNN or S3DIS.

# Training on ScanObjectNN
./train_ScanObjectNN.sh

# Training on S3DIS
./train_S3DIS.sh

Follow instructions in these scripts to change parameters.

Plotting functions

We provide plotting functions to plot the performances during and after training. They are in the plot_ScanObj.py and plot_S3DIS.py script. Here are

• Step 1: Define the dates of the logs you want to plot in the experiment functions. See example functions experiment_name_1() and experiment_name_2()

def experiment_name_1():
    ...
    start = 'Log_2020-04-22_11-52-58'
    end = 'Log_2023-07-29_12-40-27'
    ...

• Step 2: Choose the log to show

# Choose the logs to show
logs, logs_names = experiment_name_1()

• Step 3: Run the script

python3 experiments/ScanObjectNN/plot_ScanObj.py
# or
python3 experiments/S3DIS/plot_S3DIS.py

Once the trainings are finished, you can change to test mode with perform_test = True. This will start tests for the selected trained weights and show a summary of the test results.

Test models

You can also directly test a trained network with the following scripts.

# Test a model on ScanObjectNN
./test_ScanObjectNN.sh

# Test a model on S3DIS
./test_S3DIS.sh

More detailed instructions are in these scripts.

Pretrained weights

We provide the following pretrained models:

Model	Benchmark	OA	mAcc	Size	Archive
KPConvD-L	ScanObjectNN	89.7%	88.5%	80 MB	link
KPConvX-L	ScanObjectNN	89.1%	87.6%	138 MB	link

Model	Benchmark	Val mIoU	Size	Archive
KPConvD-L	S3DIS (Area5)	72.3%	151 MB	link
KPConvX-L	S3DIS (Area5)	73.5%	169 MB	link

You can download and extract them to the result folder and use our scripts to test them.