WeComm

WeCommunity Logo

Inspiration

We took inspiration from Apple's smooth, rounded design, and from OpenAI's GPT 4o, which comes with image recognition to help out WeCommunity with their drive program

What it does

It provides a way for WeCommunity to easily accept donations and run the drives safely and efficiently. It has a feature that lessens the burden on the drive organizers, by helping the donators quickly fill out what they are donating using Artificial Intelligence, as well as preventing spam with accounts. It also helps donators access the ability to donate easier because it is all online.

How we built it

We used BeautifulSoup and requests in Python to scrape the web for images of the listed donate-able items on WeCommunity, and then we trained the data using TensorFlow and hosted the backend using flask. Our frontend was built with React and Tailwind, with authentication via firebase, and it included the React Camera library, which allowed the detection of different types of items. They are then prompted to mail out their order and take a picture of their mail receipt. All of the data is then accessible for the drive organizers, and donated item counts can be easily accessed to know the scale of the drive-to-be.

Challenges we ran into

We initially started the project in pure JavaScript, but that did not work well with firebase so we restarted with react and tailwind. We also initially tried to scrape from Google, however, Google did not like being scraped so we switched to Bing. There was also a lot of bugs with the frontend, and creating the final frontend designs and implementing them took time. The AI model also has some issues that we will have to fix later, along with the flask API

Accomplishments that we're proud of

We are proud of completing this full stack project in just 12 hours, and we are also proud of successfully training our own AI model and not being tied to OpenAI's GPT or Google's Gemini. We are also very proud of helping a non profit organization help solve the homelessness community using our skills.

What we learned

We learned that training an AI model takes a lot of time, and that we need to start creating the frontend first and keep at it, and also to do perfectionism on the styling last as it is not important in the slightest in comparison to finishing the project.

What's next for WeComm

We are planning on expanding our AI drive donation detection algorithm to be real time and take in video, so that donators are able to just lay out all their items and easily add them to their donation count. We also want to re-train the AI with our own training data to make it a lot better. We would also love to get user feedback to continually improve our model and site.