edisonqkj / road-network-classification

Classification of Urban Morphology with Deep Learning: Application on Urban Vitality

This repository is the official implementation of Classification of Urban Morphology with Deep Learning: Application on Urban Vitality. It includes the major codes (written in Python) involved in the paper. We also offer some tractable tutorials in Notebook to show how to use our two modules, CRHD generator and Morphoindex generator. CRHD generator can automatically produce Colored Road Hierarchy Diagram (CRHD) for a given urban area. Morphoindex generator can automatically generate both traditional morphological indices based on built environment Shapefiles and road network class probabilities based on our road network classification model.

Requirements

To use CRHD generator, you need to install the requirements:

pip install osmnx
pip install geopandas
pip install matplotlib

To use Morphoindex generator, you need to install the additional requirements:

pip install tensorflow
pip install keras
pip install cv2
pip install numpy

If you want to use our Morphoindex generator to calculate road network class probabilities, you should also download config.py, MODEL.py and Build_model.py togehther with morphoindex_generator.py, and put them in the same filepath. Also, make sure you have downloaded our pretrained model which you can find below.

Tutorials

To let you quickly understand how to use our tools, we prepared some easy tutorials for you to have a glance:

CRHD generator tutorial

Morphoindex generator tutorial

Pre-trained Model

You can download our pretrained models here:

• Road network classification model trained on our labelled image set using ResNet-34 architecture, learning rate as 0.0005, batch size as 2.

Results

Our model achieves the following performance on the testing set:

Confusion matrix and ROC curves:

Paper

A paper about the work is available.

If you use this work in a scientific context, please cite this article.

Chen W, Wu AN, Biljecki F (2021): Classification of Urban Morphology with Deep Learning: Application on Urban Vitality. Computers, Environment and Urban Systems 90: 101706.

@article{2021_ceus_dl_morphology,
  author = {Wangyang Chen and Abraham Noah Wu and Filip Biljecki},
  doi = {10.1016/j.compenvurbsys.2021.101706},
  journal = {Computers, Environment and Urban Systems},
  pages = {101706},
  title = {Classification of Urban Morphology with Deep Learning: Application on Urban Vitality},
  url = {https://doi.org/10.1016/j.compenvurbsys.2021.101706},
  volume = {90},
  year = 2021
}

Contact

Chen Wangyang, Urban Analytics Lab, National University of Singapore, Singapore