Inspiration
- Sustainability
- Image recognition
- Automate trash and recycling separation
What it does
Our app takes a photo from either your computer or a file stored locally on your machine and uses a google Gemini pipeline to identify an object in a photo, and give advice and proper procedures on how best to recycle said item.
How we built it
We started with a brainstorming session, eliminating the majority of ideas before falling on this one. We then switched gears into a FIGMA design, determining a good basic layout, then began our implementation, first starting with resnet-50 yangy66-garbage-classification google/vit-base-patch16-224 for image recognition, but then eventually pivoted into Gemini.
Challenges we ran into
- We were generally inexperienced especially in front-end work & creating software from start to finish
- Most of the image recognition models were generally inaccurate
- Troubleshooting
- Google Gemini limit
- Project setup
- AI debugging issues
Accomplishments that we're proud of
- We finished the project on time
- We came up with an idea very quickly
- Two of us had never done a hackathon before
- We properly had a small enough idea without letting scope-creep hurt our progress
- We participated in all the events
What we learned
- Learned Figma
- Learned Node
- Machine learning
- How to interact with LLM's
- Git/version control with collaboration
What's next for TrashSort.ai
- Optimize for Industrial trash sorting
- Improve accuracy
Log in or sign up for Devpost to join the conversation.