Urban Shade Mapper

Accompanying code for the Urban Shade Mapper project for our work in Computational Learning Workshop at TAU.

This project was born out of the increasing need for shade in urban environments. High temperatures during summer months in many cities pose a major issue for the well-being of pedestrians, an issue that will only become more prevalent with global warming. Unfortunately, there are currently no good and precise tools to assess where in a city there is shade and where it is sunny. As a result, cities lack the necessary information to invest their resources efficiently into improving the shade situation.

To address this issue, the project aims to develop a digital tool for shade optimization. This tool will enable the planning of solutions for shade provision at the street level, ultimately improving the well-being of the city's residents. We achieve this using street-level imagery data and machine learning models that helps us assessing the shade situation on the street.

Installation

1. Download sam_vit_h_4b8939 model checkpoint from here and place it in the shade_segment_anything folder.
2. Start a virtual environment using venv:

python3 -m venv venv
source venv/bin/activate

3. Install requirements:

pip install -r requirements.txt

Input

The program expects to receive a file named test_points.csv. The CSV file should be in the following format:

name,lat,lon,date_of_interest,pano_id,short_intervals
1,32.0916058,34.77306685,2023-07-18,_,Y
2,32.0915486,34.7733054,2023-07-18,_,N
3,32.0914841,34.7735563,2023-07-18,_,Y

where each column represents:

• name: index of the point
• lat : latitude coordinate
• lon : longitude coordinate
• date_of_interest: date of simulation in format YYYY-MM-DD
• pano_id: GSV panorama id (can be left blank)
• short_intervals: 'Y' for prediction every 10 mins, 'N' for every hour

The test_points.csv file should be in the same directory as main.py

Running

Just run the following command:

python main.py

Output

The outcome will be stored in a CSV file with following columns:

• point: index of the point
• time: time of day
• is_shade: shade status, 1 for yes, 0.5 for borderline, 0 for no
• building_pixels/tree_pixels/sky_pixels: percentage of pixels in the frame which were classified as building/tree/sky
• nearest_pano_coordinates: coordinates of the location we did compute from.

Please note: due to the many hours in a day, point, tree_pixels, building_pixels, sky_pixels, nearest_pano_coordinates are all the same for a given point, only time andis_shade values change.

The output file will look something like this:

point, time, is_shade, tree_pixels, building_pixels, sky_pixels, nearest_pano_coordinates
1, 08:00, 1, 45, 30, 25, 32.0916058,34.77306685
1, 08:10, 0.5, 45, 30, 25, 32.0916058,34.77306685
1, 08:20, 0, 45, 30, 25, 32.0916058,34.77306685

Remarks

This tool provides an efficient way to predict the presence of shade for a given geographical point across different times of the day. Always ensure that your input file is in the correct format for the best results. Happy shading!

Please open an issue if you encounter any bugs or have suggestions for further improvement. Contributions to improve this project are welcome.