Shape Classification of Building Information Models using Neural Networks

Overview

This repository contains the source code of the short paper presented in 3DOR workshop 2021. Please find below links to

Author version of the paper
Video presentation

Repository contents

The project is composed of the following parts :

An implementation of the solid voxelization algorithm
The neural network architecture
A dataset folder with a template for each type of BIM class

Solid Voxelization

A single file implementation of the solid voxelization algorithm. The main source code can be found in :

voxelizer\src\voxelizer.cpp

Building and running the code will require :

Visual Studio 2017 or 2019

Neural Network Architecture

A tensorflow 2.4 implementation of the proposed neural architecture can be found in :

python\src\model.py

An example entry point for training and testing our model can be found in :

python\train.py
python\test.py

The augmentation, normalization and alignment function implementations are located in :

python\src\utils.py
python\src\dataset_builder.py

An example entry point for creating the dataset is located in :

python\preprocessor.py

Dataset

In python\dataset\BIM templates directory, one can find the 3D model templates for each one of the 16 classes used for evaluation in the paper.

Shape Classification of Building Information Models using Neural Networks
I. Evangelou, N. Vitsas, G. Papaioannou, M. Georgioudakis, A. Chatzisymeon
to appear: Eurographics Workshop on 3D Object Retrieval (3DOR) 2021 short papers [ Paper ]