Inspiration
One of our dedicated team members volunteered at a local food bank and noticed a significant lack of participation from the community. Inspired by this observation, we embarked on a mission to make food donation more engaging and rewarding.
To achieve this, we gamified the food bank experience by introducing a leaderboard and ranking system. This allows volunteers and donors to compete in a friendly manner while contributing to a noble cause. Participants can create profiles to showcase their contributions, track their progress, and view the food items submitted by others.
What it does
FoodBank AI uses computer vision to recognize all sorts of foods that can be donated to a food bank and will give you a score based on how non-perishable and nutritious it is. Users are able to compete with each other to get the highest score and try to be ranked #1 on the leaderboard!
How we built it
We split up the tasks amongst each other to make things more efficient. We used ResNet as a pretrained model to fine-tune food images, which would help to increase accuracy for food recognition. On the backend we used PostgreSQL and Flask, for the frontend, we used React and Figma. We tested the model by taking pictures of fruit from our webcams. We also used an API, FoodData Central to get the nutrient facts for a food item that was recognized by our model and then we made a formula to spit out a score, which would be used in the leaderboard.
Challenges we ran into
FoodBank AI uses computer vision which none of us had any experience in, but nevertheless, we took the opportunity to learn and make a project out of it. We also ran into a lot of merge conflicts, to solve this we went into our own branches and were able to resolve any conflicts that we may have had.
Accomplishments that we're proud of
Being able to finally see the model able to make a predication of what the food is from a photo taken from our laptops was incredibly satisfying and we would see that as a big accomplishment. Considering how little time we had to start and complete this project, the fact that we had a finished product was also a massive accomplishment.
What we learned
- How to preprocess and clean images as data for a computer vision model
- Learned React
- How to finetune a pretrained AI model
- Learned how to create a Flask backend
What's next for FoodBank AI
- Curate more data for our model to increase accuracy
- Make the UI more visually appealing
- Add user authentication to our application through Firebase
Log in or sign up for Devpost to join the conversation.