BottleBot

• 

  Home screen!

Inspiration

Meet Josh. It is his first hackathon. During his first hack, he notices that there is a lot of litter around his work area. From cans to water bottles to everything else in between, there was a lot of garbage. He realizes that hackathons are a great place to create something that can impact and change people, the world, or even themselves. So why not go to the source and help hackers be that change?

What it does

Our solution is to create a mobile application that leverages the benefits of AI and Machine Learning (using Google Vision AI from Google Cloud) to enable the user to scan items and let the user know which bin (i.e. Blue, Green or Black) to sort the item to, as well as providing more information on the item.

However, this is not merely a simple item scanner. We also gamified the design to create a more encouraging experience for hackers to be more environmentally friendly during a hackathon. We saw that by incentivizing hackers with awards such as "Earthcoins" whenever they pick up and recycle their litter, they can redeem these "coins" (or credits) for things like limited edition stickers, food (Red Bull anyone?) or more swag. Ultimately, our goal is to create a collaborative and clean space for hackers to more effectively interact with each other, creating relationships and helping each other out, while also maintaining a better environment.

How we built it

We first prototyped the solution in Adobe XD. Then, we attempted to implement the idea with Android Studio and Kotlin for the main app, as well as Google Vision AI for the Machine Learning models.

Challenges we ran into

We ran into a lot of challenges. We wanted to train our own AI by feeding it images of various things such as coffee cups, coffee lids, and things that are recyclable and compostable. Although, we realized that training the AI would be a lengthy process. Another challenge was to efficiently gathering images to feed into the AI, we used Python Selenium to automate this but it required a lot of coding. In order to increase the success rate of the AI identifying certain things, we would have to take our own pictures to train it. With this in mind, we quickly shifted to Google Cloud's Vision AI API, though we got a rough version working but was not able to code everything in time to make a prototype.

One more challenge was the coding aspect of the project. Our coder had problems converting Java code into Kotlin, in addition to integrating the Google Cloud Vision API into the app via Android Studio.

We had further challenges with the idea that we had. How do we incentivize those who don't care much about the environment to use this app? What is the motivation to do so? We had to answer these questions and eventually used the idea of hackers wanting to create change.

Accomplishments that we're proud of

We're proud of each other for attempting to build such an idea from scratch, especially since this was their first hackathon for two of our team members. Trying to build an app using AI and training it ourselves is a big idea to tackle, considering our limited exposure to machine learning and unfamiliarity with new languages. We would say that our accomplishment of creating an actual product, although it may have been incomplete, was a significant achievement within this hack.

What we learned

We gained a lot of first-hand insight into how machine learning is complex and takes a while to implement. We learned that building an app with external APIs such as Google Vision AI can be difficult to do compared to simply creating a standalone app. We also learned how to automate web browser tasks with Python Selenium so that we could be much more efficient with training our AI.

The most important thing that we learned was from our mentors was regarding the "meta" of a hackathon. We learnt that we have to always seriously consider our audience, the scope of the problem, and the feasibility of the solution. The usability, motivation, and the design are all major factors that we realized are game changers as one certain thing can completely overturn our idea. We gathered a lot of insight from our mentors from their past experiences and w are inspired to use what we learned through DeltaHacks 6 in other hackathons.

What's next for BottleBot

The aim for BottleBot's future is to fully integrate the Google Cloud Vision API into the app, as well as to finish and polish the app in Android Studio.

Built With

• google-cloud
• google-vision-ai
• kotlin
• python
• selenium