IntelliCycle

• 

  InteliCycle

Inspiration

Our planet protects us from harmful radiation, and all the dangers of space, as well as provide us with vital resources needed for the survival and development of humanity. However, we often neglect our planet by producing more and more waste products.

Recycling is one of the most impacting ways we can help with sustainability and the ecosystem. Most people want to recycle, however, the process is often too difficult which pushes people away from it. That’s why we built this application to make this process simpler for people who want to recycle.

Doing so will change the way the World currently looks by saving energy, significantly reducing landfill waste, and protecting our environment.

What it does

Our mobile application allows the users to capture an object with their camera, educates about the type of waste, and displays nearby locations where they can recycle the specific object.

How we built it

We built our front end using ionic with vue.js, this will allow our app to be a multiplatform app by changing just a few settings for IOS and/or Android. The backend is powered by machine learning, we apply few-shot learning using K-nearest neighbor search on image embeddings from OpenAI CLIP vision transformers. With our approach, we can scale our image classifier with little data per class without fine-tuning a new model.

Challenges we ran into

Connecting our API backend and other APIs (Google Maps, Recycling Directory) together was the most challenging part. Additionally, we had issues with getting started with the model due to version and CUDA conflicts.

Accomplishments that we're proud of

We are proud of the accuracy of the model. With just a small dataset we have currently in the backend GitHub, we can classify new images with high accuracy. We are happy with the app and hope that it will a used in the future.

What we learned

We increased our knowledge about the types of recyclable waste, and how important it is, as well as deepen our understanding of machine learning and its possibilities. Moreover, we learned how to develop an application in a very limited amount of time while cooperating with technologies like GitHub, google docs, slack, etc.

What's next for IntelliCycle

See github page at

https://github.com/grosso379/recycling-ui https://github.com/OutWrest/recycling_model

Built With

• fastapi
• ionic
• pytorch
• vuejs