Hearing the theme of space we decided that we wanted to make something with satellite data. Also, we were interested in doing a hack for social good. Combining the two we wanted to analyze the spread of different diseases. This led us to the Malaria Atlas Project as it had ample amounts of data.
What it does
This tool allows for researchers to analyze malaria hotspots and how they correlate with other factors. Using our model, they will be able to view locations where a malaria outbreak is most likely to occur. The way it does this is it starts by allowing the researcher to either enter a country or continent they are interested in studying. From there, they have the option to add their own observed variables for that year. If they choose to do that, we will then provide them a predicted model based on their inputs. Otherwise, we will just show them our model and allow for them to interact with it in 3D.
How I built it
To build this app, we started by pulling all the data from the Malaria Atlas Project. This was done both with R plumber and Python with OWSLib. From there we began to analyze the data. To do this we began looking into the state of the art models. The decision then came to either using a GAN vs using an LSTM. Training on a rather large multivariable set of data, we decided to implement a GAN. The GAN was made using Python and to allow for the script to communicate with the researcher we used Flask to connect the data to the researcher. Lastly, to render the actual map, we decided to use WebGL and then also, because we had a VR set, we decided to map it onto a 3D map.
Challenges I ran into
The biggest issues we ran into were downloading the data and using the VR set.
Accomplishments that I'm proud of
Built a better API than MAP. Created a predictive model for Malaria using Generative Adversarial Networks (GANs).
What I learned
How to use an oculus, how to use a GAN to predict frames.
What's next for MAIPLE
Hope to improve our model with crowd-sourced data.