llien30 / point_cloud_anomaly_detection

TOWARD UNSUPERVISED 3D POINT CLOUD ANOMALY DETECTION USING VARIATIONAL AUTOENCODER

(accepted ICIP2021)

Usage

Before start training, please setup earth mover's distance module (See libs/emd/README.md)

Make environment

conda env create -f=emd_env.yml

Compile

cd libs/emd
python setup.py install

to check the compile is success, please run

python emd_module.py

after the last line of emd_module.py(test_emd()).

Training

python train.py ./config/[config file name] (--no_wandb)