Our inspiration for this project came from the recent riots in our city. The streets of our city were defaced and we realized that there had to be a better alternative than driving around the city aimlessly. We came up with the WeClean app to help assist our community in intuitively and efficiently locating places to clean. That way, our community post-cleanup will be as beautiful as before.

What it does

Users upload photos of damaged areas. WeClean adds these images to the server, as well as its location, allowing other users to help restore the area. This way, users can help clean when and where they can.

Challenges we ran into

This was our first time working with TensorFlow; it took us more than five hours and two developers to get both iOS and Android working, and after it finally worked, we spent 16 whole seconds celebrating before starting the next thing. Additionally, configuring OAuth caused us to pull half our hair out, as iOS and Android require different configurations. Configuring our Google Places and Reverse Geocoding api was also a challenge, as we had never written http requests to outbound URLS with NodeJS before. Finally, we had to manually locate and tag numerous unique photos of graffiti on murals, brick walls, and broken glass in order to train our AI, and give it enough time to build a model.

Accomplishments that we’re proud of

We are proud that we have created a cross platform app with both a dark and light mode that uses native widgets (such as Google Maps and Apple Maps) for each platform for a native feel. Additionally, we managed to gather data to train a TensorFlow model with Firebase ML using our own dataset with 274 labeled images to train the AI. We are also proud to offer OAuth Google sign-ins. In addition, we have cloud functions that use Google Maps reverse geocoding.

What we learned

We learned two new programs: TensorFlow & To make a custom TensorFlow model, we used Firebase Auto ML. This was unlike anything we’d done in a hackathon before, and was a fun learning experience. We tried in a previous hackathon to get working, with no luck. This time around, with a fresh take and more experience, we finished it.

What's next for WeClean

WeClean will grow with a more intuitive way to add more cleanups, slightly improved UI/UX, adding a dark themed Google Maps for Android, adding more ways to sign in, and, most importantly, adding more data to our ML model.

Built With

Share this project: