WasteX AI

"AI for smart waste management"

Comment

WasteX AI

Inspiration

Living in rapidly growing urban areas, we noticed a recurring problem: overflowing waste bins. They aren't just an eyesore; they are a public health hazard and a source of pollution. We realized that the current waste management cycle is reactive—garbage trucks follow fixed schedules regardless of whether a bin is empty or overflowing.

We asked ourselves: What if we could give the city "eyes"? Inspired by the capabilities of multimodal AI, we set out to build WasteX AI, a platform that transforms simple images into actionable data for smarter, cleaner cities.

What it does

WasteX AI is a real-time smart city dashboard that automates waste monitoring.

  1. Snap & Analyze: Users upload a photo of a waste bin.
  2. Gemini Intelligence: The app uses Google's Gemini 2.5 Flash model to visually analyze the image and determine the "fill level" (Low, Medium, High, or Critical).
  3. Live Mapping: The data is geotagged and displayed on an interactive map, allowing municipal authorities to identify hotspots instantly.
  4. Optimized Routing: By prioritizing "Critical" bins, cleanup times can be reduced by up to 50%.

How we built it

We built WasteX AI as a full-stack application using the Next.js 16 App Router for a seamless frontend and backend experience.

  • AI Engine: We integrated the @google/generative-ai SDK. The core analysis happens via a carefully crafted prompt sent to the gemini-2.5-flash model, which accepts the image as input and robustly classifies the waste level.
  • Database: We used MongoDB with Mongoose to store report data, specifically leveraging geospatial queries to map the latitude and longitude of every report.
  • Frontend & Visualization:
    • Tailwind CSS & Geist UI principles for a clean, modern aesthetic.
    • React Leaflet for the interactive map that renders live waste markers.
    • Recharts for visualizing data trends in the dashboard.
  • Authentication: Secure user management to ensure trustworthy reporting.

Challenges we faced

  • Prompt Engineering for Vision: Clear distinction between a "full bin" and a "large bin" was tricky. We had to iterate on our system prompt to ensure Gemini focused specifically on the trash level inside the bin, not just the bin's physical size.
  • Real-time Latency: Initially, image upload and analysis took too long. Switching to Gemini 2.5 Flash significantly reduced inference time, making the app feel snappy and responsive.
  • Map Integration: Handling client-side map rendering (Leaflet) within Next.js Server Components required careful state management and dynamic imports.

Accomplishments that we're proud of

  • Seamless AI Integration: We successfully built a pipeline where a raw image from a phone is converted into structured database capability in under 2 seconds.
  • UI/UX Design: Creating a dashboard that doesn't just look like a spreadsheet but feels like a modern mission control center.
  • Impact Potential: Realizing that this tool could genuinely help sanitation workers prioritize their routes and save manual effort.

What we learned

  • Multimodal AI is accessible: You don't need to train a custom TensorFlow model for weeks. Gemini's out-of-the-box vision capabilities are incredibly powerful for real-world object analysis.
  • The power of Next.js 16: Using the latest features of Next.js helped us move fast, keeping our API and UI logic tightly coupled but clean.

Built With

Share this project:

Updates