Inspiration

We were inspired by the catastrophic consequences that COVID-19 has had on society. There have been massive socio-economic ramifications for billions of individuals around the world, inciting consumer fear within many. People are afraid to even step foot outside and it’s extremely unhealthy. It is critical that we smoothly integrate everyone back into the norms of society while also safely providing a solution to inform an individual whether a public location is deemed safe. Additionally, the planet has seen great improvements in carbon emissions because of COVID-19, so we wanted to continue the effort to stay green and protect our planet.

What it does

We made a mobile app that informs a user whether they should travel to a chosen location. The user types in a location that he/she would like to travel to. The app then tracks the number of users within that place. Using this information, the app calculates the density of the people within the place based on how many people can fit in the location safely under World Health Organization guidelines. This data is then displayed as a heat map that tells the user whether they should make their journey. Additionally, the app finds the greenest ways to travel in order to be environmentally friendly.

How we built it

The app portion of the software was built completely in React Native. We used this framework primarily for its cross-compatibility functionality. The app uses modules such as the Google Maps API and the GeoSpark SDK to map heatmaps and track the user's location. In regard to the API, we used Tensorflow/Keras to develop a recurrent neural network that would be able to generate a heatmap based on time series data that we provided to it, in order to predict future population patterns for a specific area. In order to insure privacy for all users, we were able to develop a blockchain by using IBM’s baseline blockchain python classes and modifying them for our own use. Finally, we deployed this API using Flask, which enabled us to actually provide our services in the form of a REST API, and we decided to use Docker for portability/scalability after seeing the talk on containerization. Finally, we integrated the core app with this API so the app was able to use its functionalities.

Challenges we ran into

We were challenged on how to predict more than one step into the future for the LSTM model. Initially, an iterative method of continually plugging in the output of the model as the new input was used, but this did not preserve accuracy well enough to be viable over predictions further into the future. The issue was resolved by switching the model from predicting based on the most recent data point to training the model to predict the next k values based on the previous k values. We also had a few bumps along the road while integrating GeoSpark with the app and with inaccurate heatmap renderings. We fixed this by adjusting the settings and options to optimize the heatmap data.

Accomplishments that we're proud of

We’re extremely proud that we were able to build an app that can immediately help millions and help this beautiful planet. We feel that our biggest achievement was being able to integrate so many different technologies and platforms in such a short time period while still retaining the overall functionality of the product.

What we learned

We learned how to utilize and implement many new technologies. We learned about the GeoSpark API, and we also greatly expanded our knowledge about how BlockChain can be used to protect the anonymity of a user in a network. Additionally, we also learned about Recurrent Neural Networks and how they are beneficial in evaluating time series data to offer future insight.

What's next for SafeSpot

The biggest next step for SafeSpot is to work with telecommunication companies in order to gain substantial amounts of consumer data. Currently, the data was created by us for prototyping purposes, meaning it has not been verified. By working with these companies, we can easily gain large amounts of data to proficiently train our model and improve the reliability of our product while also being professionally validated.

Built With

  • containerization
  • docker
  • flash
  • geospark
  • geospark-for-location-tracking
  • keras
  • numpy
  • python
  • python-for-the-blockchain-technology
  • react-native
  • tensorflow
  • tensorflow/keras-and-numpy-for-machine-learning-implementations
Share this project:

Updates