Workshop about Python for Urban Analytics as part of the Singapore-ETH Centre (SEC) activities for the National Coding Week 2024.

• Have an initial understanding of Street-View imagery use cases and data manipulation
• Overview of open source resources SEC and collaborators are producing

The slides used for this workshop can be found in this repo in the file 20240916-python-workshop.pdf.

Install Python 3.10.14 and run

pip install -r requirements.txt

You may need to double check for the geos package. On a mac, run

conda install -c conda-forge geos

Create a Mapillary account and get an API token.

We will use the Global Streetscapes dataset! With more than 9 million images from 688 cities and 300+ features, it's the right place to start playing with SVIs.

• Project
• Paper
• Dataset
• Code and wiki

As mentioned in the Download Images wiki, we will use the code from code/download_imgs.

Go to this URL and paste the folder URL.

Unzip the folder within this repo and rename the folder to download_imgs/.

As mentioned in the Dataset card, please avoid using git clone to download the repo as Git stores the files twice and will double the disk space usage to 124+ GB.

Make sure you have at least 40GB of free space.

Run the file 1-download-huggingface.py to download the data/ folder as well as the additional files cities688.csv and info.csv.

Then, follow the notebook 2-data_download.ipynb to generate the final .csv with the desired images.

Finally, as mentioned in the Download Images wiki, go to download/imgs/download_jpegs.py and update the files paths:

in_csvPath = 'download_imgs/selected_points.csv' # input csv, the name of the saved file in 2-data_download.ipynb
out_mainFolder = './global-streetscapes/svis/'

In this file, and in download_imgs/download_jpegs_mapillary.py, update your Mapillary token. Finally, run:

python download_imgs/download_jpegs.py

The notebook for this is hosted on kaggle: https://www.kaggle.com/code/matiasqr/semantic-segmentation

The script 3_inference.py is the same codebase used in the Global Streetscapes.

Run python 3_inference.py

The notebook 4_perception-comparison.ipynb elaborates on the different city perceptions.

• Hou, Y., Quintana, M., Khomiakov, M., Yap, W., Ouyang, J., Ito, K., Wang, Z., Zhao, T. & Biljecki, F. (2024). Global Streetscapes — A comprehensive dataset of 10 million street-level images across 688 cities for urban science and analytics. ISPRS Journal of Photogrammetry and Remote Sensing, 215, 216–238. https://doi.org/10.1016/j.isprsjprs.2024.06.023