Eco Lens

Home Page

Inspiration

We noticed that throwing garbage was no simple task, we found different bins for different waste materials around campus. In one instance it would take us about 5 mins to figure out where to throw a particular item. From this we realized that recycling is a challenge. After a bit more research we found out that the way garbage is recycled is still done mostly manually and using expensive equipment like strong magnets and lasers. There is an opportunity to use AI based image classification models to accurately categorize these items in such centers, and feeding these results into a robotic arm in real time could help sort waste materials efficiently. Using this as an inspiration we created Eco Lens.

What it does

In it's current stage of development we developed a mobile web application that anyone can use to classify given objects either by uploading an image or grabbing a picture through your camera. The AI image detection model detects what sort of object it is and we pass this on to an AI text classifier which will determine what category the object falls into (example: plastic bag, falls into the non-recyclable category). Additionally the classifier will also provide information on how an average person can recycle it themselves.

How we built it

We developed the application on the React.js framework, utilizing the Vit model from Hugging Face to identify the image and using the Llama model from Groq for the text generation. We set up a collaborative development environment on Github and hosted the website app on a domain from GoDaddy using Github Pages.

Challenges we ran into

All three members of our team are participating in our first hackathon, utilizing AI models for the first time. We ran into challenges attempting to train our own AI model as such we opted to using a pre-train AI model.

Accomplishments that we're proud of

For the first we utilized AI models into our applications. We also managed to create a collaborative development environment as well as creating a deployment workflow, all within 2 days.

What we learned

We learned to utilize AI models and integrating API's to web applications as well as working as a team, developing a fully functional prototype application within 48 hours; with a polished, elegant and user friendly UI.

What's next for Eco Lens

In regards to assisting recycling centers the next steps are:

Train our model to increase accuracy.
Feed data into a robotic accessory to enable sorting of waste.
Try to get real time data by getting access to a recycling center to generate a training dataset. In regards to the mobile application available to general users:
The App should also be able to provide locations of relevant disposal centers, example:
- Object is laptop, location of electronic disposal center
- Object is clothing, location of thrift stores to recycle, or local recycling centers.