GreenGuard

Inspiration

It is known that 70% of recyclable and compostable waste is in landfills due to a lack of proper sorting and accessibility to disposal facilities. We realized that by leveraging AI and location-based technologies, we could bridge this gap and help people properly classify waste and find the nearest disposal facilities.

What it does

GreenGuard is an AI-powered waste classification app that allows users to snap a photo of any waste item, and our TensorFlow model classifies it into one of five categories: Recyclable, Compostable, Textile, E-Waste, or Hazardous. It then uses Geoapify and EPA Envirofacts API to recommend nearby disposal facilities for the detected waste type.

How we built it

Data Collection: Sourced from a combination of Kaggle's waste classification datasets. Model Training: Built a CNN model using TensorFlow for waste recognition. UI Development: Designed a cross-platform app with Kivy. Location Services: Integrated Google Geolocation API and Google Maps Geocode API.

Challenges we ran into

We encountered challenges such as limited dataset diversity, which required data augmentation. We also had to repeatedly fine-tune the model to distinguish between similar materials like plastic and glass. This was inefficient as it took at least one hour each time to train the model. Additionally, Google API rate limits for real-time facility recommendations were another massive challenge.

Accomplishments that we're proud of

• Successfully achieving high-accuracy waste classification.
• Seamless real-time mapping of nearby disposal facilities.
• Building an end-to-end AI pipeline within 24 hours.
• Enhancing user awareness and sustainable behavior.

What we learned

We learned the importance of data preprocessing and augmentation for better model accuracy. We also learned how to integrate AI with geolocation services for real-world impact. Additionally, we learned effective team collaboration and time management in a high-pressure environment.

What's next for GreenGuard

• Expanding our dataset for improved classification accuracy.
• Adding community reporting features for illegal dumping sites.
• Implementing a rewards system to encourage eco-friendly habits.
• Partnering with local recycling facilities and municipalities for broader impact.
• Expand the platforms list and make the app available on mobile devices.

Built With

• google-geolocation-api
• google-maps-geocode-api
• kaggle
• kivy
• python
• tensorflow