Knowing SafeTrek's intentions to make 911 calls obsolete and incorporate various technologies to communicate exactly what is going on at the scene swiftly and concisely, we decided to head in that direction and use smartphone camera capability and Machine Learning to quickly communicate to the proper authorities what the situation is and where it is.

What it does

Users will open the app, select the large camera icon and be immediately taken to the camera screen. Once the user takes a picture, they have the option to use it or retake it. Once they decide to use it, our ML models determines what type of emergency it is and alerts the specific authorities that need to be at the scene and notifies the user that help is on the way.

How we built it

We used React Native for our mobile based front end. For our back end, we used python. Specifically, for our Machine Learning model we used the Keras library that has a TensorFlow backend. To connect the two ends, we used a server hosted by Amazon Web Services. We finally used the SafeTrek API to send an alert to specific authorities such as the fire department or police department based on which ones have to be present at the scene.

Challenges we ran into

Neejole and Ethan, who were in charge of the front end, had NO experience whatsoever with JavaScript, much less React Native itself. It was definitely a challenge for them to learn it, however, the React Native workshop greatly reduced the start up time. Furthermore, Swaraj and Sean, though both have had some previous knowledge of Machine Learning, have never used Keras to implement it. This also posed many issues, including an issue with TensorFlow and multi-thread environments.

Accomplishments that we're proud of

Everyone on the team learned something new to add to their toolset and we were very proud of our final product. One main thing we're proud of was getting the server with AWS to work properly and efficiently combine the two ends.

What's next for SafePhoto

The future of SafePhoto would involve the photo that the user takes being sent to the authorities to be notified.

Built With

Share this project: