IFCNet

Update 13/06/2022: The code for our follow-up paper SpaRSE-BIM: Classification of IFC-based geometry via sparse convolutional neural networks will be added soon.

Neural Network Models for the IFCNet Dataset. The paper can be found here.

Installation and Training

Make sure you have PyTorch installed. The code in this repository was developed and tested with:

• Python 3.8.5
• PyTorch 1.7.1+cu110
• Ubuntu 20.04

git clone https://github.com/cemunds/ifcnet-models.git
cd ifcnet-models

It is recommended to create a new virtual environment before installing the dependencies with the following command:

pip install -r requirements.txt

Download the data for the model you want to train and place them in the corresponding data folder:

• MVCNN
• DGCNN
• MeshNet

For example, for MVCNN:

mkdir -p data/processed/MVCNN
cd data/processed/MVCNN
wget https://ifcnet.e3d.rwth-aachen.de/static/IFCNetCorePng.7z
7z x IFCNetCorePng.7z

Navigate back to the root directory of the project and execute the training script.

cd ../../..
python src/models/train_model.py MVCNN

Running the training script like this will perform a hyperparameter search using Ray Tune and Optuna. If you want to use a fixed set of hyperparameters, you can specify the path to a configuration JSON file with the --config_file flag. For the format of the configuration file, please refer to the files included with the pre-trained models.

Warning: Performing the hyperparameter search for MVCNN consumes a lot of disk space, as currently the model is saved to disk after every epoch and some of the backbone models (e.g. vgg11) are very large. I intend to change the saving behavior in the future.

Evaluation

The pre-trained models can be downloaded here:

• MVCNN
• DGCNN
• MeshNet

Download and unzip the models into the models directory:

mkdir models
cd models
wget https://ifcnet.e3d.rwth-aachen.de/static/mvcnn_model.7z
7z x mvcnn_model.7z

Now you can start a Jupyter Notebook server and run the notebook for the corresponding model.

Citation

If you use the IFCNet dataset or code please cite:

@inproceedings{emunds2021ifcnet,
  title={IFCNet: A Benchmark Dataset for IFC Entity Classification},
  author={Emunds, Christoph and Pauen, Nicolas and Richter, Veronika and Frisch, Jérôme and van Treeck, Christoph},
  booktitle = {Proceedings of the 28th International Workshop on Intelligent Computing in Engineering (EG-ICE)},
  year={2021},
  month={June},
  day={30}
}

Acknowledgements

The code for the neural networks is based on the implementations of the original publications, but had to be updated for recent PyTorch versions:

• WangYueFt for DGCNN
• jongchyisu for MVCNN
• iMoonLab for MeshNet

Big thanks goes to xeolabs for the xeokit SDK, which is used in the viewer.

The structure of this repository is loosely based on the cookiecutter data science project template