Welcome to the SatNet disease risk prediction software. This program works to provide a relatively inexpensive method of determining locations within cities that at the highest risk of disease outbreaks. In order to accomplish this, I utilized publically available satellite images uisng the google maps API as well as a novel hybrid residual CNN model, which is comprised of a ResNet-50, MobilNetV2, and VGG-16, in order to predict locations within cities that are most suceptible to disease outbreaks. The model is able to classify different regions as slums (poor), industrial, or low-density high-income residential which allows specific regions to be identified as danger zones. Furthermore, the SatNet software uses data from the World Bank API and the classification of city regions in order to generate a general disease risk index which highlights the overall danger the city faces from a disease outbreak when compared to cities across the world. My overall goal with this project was to create an inexpensive tool that people and governments across the world can use in order to identify disease hotspots where resources can be allocated to mitigate the spread of communicable diseases. The recent devestation caused by Covid-19 has shown the importance of early detection softwares as the impact the pandemic rocked the world. However, poorer countries struggle achieve early detection due to immense resources that are required for branches like the CDC and WHO. My hope is that SatNet can provide a cheap alternative for struggling countries so they can protect their citizens as effectively as everyone else!
The technical components of SatNet can be seperated into 3 main groups: satellite image collection, classification model, and the data visualization. The first step of the process is the collection of satellite images which is done through the google maps API. When the user types in the city they want to view, the GeoCode python library is used to find its coordinates and then a total of 400 evenly spaced, 640x640 resolution images are collected in a 5 mile radius around this point. In order to make these API calls, the google maps API instructions are followed as shown on the google maps website:
These satellite images are then inputted into the classification algorithm which is a hybrid model consisting of a ResNet-50, MobilNetV2, and VGG-16 that has been modified to include more residual components. This final layer of the model contains a sigmoid activation function which indicates the probability of each image being a slum, industrial sector, or up-scale community. Finaly, using these outputs, a heat map is generated, where red squares indicate poverty, blue indicates industry, and green indicates wealth, which the user can use to determine disease hotspots within the city.
- Download the repository as a zip folder. Extract the entire folder and save all contents into one main folder. Create a new folder in the project directory named "Images"
- Make sure to have 3.10.9 downloaded: https://www.python.org/downloads/release/python-3109/
- Open up the command prompt and enter the directory where the files have been stored. Type "pip install -r libraries.txt" into the command prompt to download all of the necessary python libraries.
- Visit the following link and download the file: https://drive.google.com/file/d/1tjqo3YFVMUxOcH1sUHr3dktlfb3AXUFz/view?usp=sharing. Save the file into the same folder as all other files.
- Visit the google maps api to generate a personal API key. Once generated, copy this key and paste it into the dictionary named "important_info" inside of the quotation marks that come after "Gmaps_api_key".
- Visit the geocode api to generate a personal API key. Once generated, copy this key and paste it into the dictionary named "important_info" inside of the quotation marks that come after "Geocode_api_key".
- Run the Test.py file and wait for the prompt to type in a city. Type in any city in South America or South East Asia and wait for 8-10 minutes for the program to run.
- The final output should be a matplotlib window that contains the image of the city to the left and the heatmap output on the right:
- The code is still in development phase and so the instructions are quite rudimentary. If you have any questions or want a more detailed set of instructions, please email me at parkiratsandhu1@gmail.com
