Theme: Social Impact
Problem statement (AI for Social Impact):
Civic waste management relies heavily on manual reporting and inefficient collection routes, leading to unverified garbage dumping and overflowing municipal dustbins. Furthermore, citizens lack the motivation and transparency to actively participate in keeping their communities clean.
CleanSweep
CleanSweep is an end-to-end “AI for Social Impact”, multi-module platform designed to be the central nervous system for a city's waste management. It transforms civic waste reporting from a manual, unverified chore into a gamified, data-driven ecosystem.
It serves as a civic waste management platform that empowers citizens to report garbage, track cleanups, earn rewards, and build cleaner communities — all in one app.
Powered by Wolfram
This project leverages Wolfram Machine Learning to automatically classify and verify images of waste, ensuring accurate and efficient civic reporting.
- 📍 Report Garbage — Submit waste reports with photo, location pin, severity, and an optional note
- 🗺️ Explore Map — Browse all reports on an interactive map with status filters
- 🏘️ Community Feed — See what others are reporting, like posts, and join discussions
- 🏆 Points & Leaderboard — Earn points for logging in, reporting waste, and confirming cleanups
- 🛍️ Marketplace — Redeem points for rewards and offers. (Rewards are funded through municipal partnerships, local business sponsorships, and CSR initiatives aimed at community cleanliness).
- 👤 User Profile — View your stats, report history, and earned badges
- 🛠️ Admin Dashboard — Municipal staff can manage reports and update dustbin statuses
| Layer | Technology |
|---|---|
| Frontend | React 19 + Vite |
| AI / ML | Wolfram (Cloud API) |
| Styling | Vanilla CSS (Glassmorphism dark theme) |
| Backend | Supabase (Auth, Database, Storage) |
| Maps | Leaflet.js |
| State | Zustand |
| Routing | React Router v7 |
- Node.js 18+
- A Supabase project
# Clone the repo
git clone https://github.com/githubutsav/CleanSweepV1.git
cd CleanSweepV1
# Install dependencies
npm installCreate a .env file in the root with your credentials:
VITE_SUPABASE_URL=your_supabase_project_url
VITE_SUPABASE_ANON_KEY=your_supabase_anon_key
VITE_WOLFRAM_API_URL=your_wolfram_classify_api_url
VITE_WOLFRAM_ROUTE_API_URL=your_wolfram_route_api_urlnpm run devOpen http://localhost:5173 in your browser.
src/
├── pages/
│ ├── Home.jsx # Main app (report, map, community, leaderboard)
│ ├── Login.jsx # Auth page
│ ├── Profile.jsx # User profile & stats
│ ├── Marketplace.jsx # Points redemption store
│ └── Admin.jsx # Municipal admin panel + Route Optimizer
├── lib/
│ ├── supabaseClient.js # Supabase instance
│ └── store.js # Zustand global state
└── index.css # Global styles & design tokens
wolfram/
├── classify_waste.wl # AI image classification (ImageIdentify)
├── optimize_route.wl # TSP route optimization (FindShortestTour)
└── README.md # Wolfram deployment instructions
The app uses Supabase with the following core tables:
profiles— User info, points, and statsgarbage_reports— Citizen-submitted waste reportsdustbin_status— Municipal dustbin trackingcommunity_posts— Community feed postspost_likes/post_comments— Engagement on postsleaderboard— Top contributors view
Note: SQL setup scripts are located in the
database/directory. Rundatabase/setup.sqlin your Supabase SQL editor to create the necessary tables and policies.
CleanSweep uses Wolfram as its computational intelligence layer, deployed as REST APIs on Wolfram Cloud.
When a citizen photographs waste, the image is sent to a Wolfram Cloud API that uses neural networks and Wolfram's Knowledge Base to return rich structured data:
| Property | Wolfram Function | Example |
|---|---|---|
| Category | ImageIdentify |
"Plastic" |
| Severity | Custom Classify |
"High" (score 3/5) |
| Recyclable | Knowledge Base lookup | ♻️ Yes |
| Decomposition | EntityValue |
"450 years" |
Scalability Note & Fallback: To ensure high availability and robust performance, CleanSweep is designed to fallback to vision models like Google Gemini if the Wolfram Cloud API encounters rate limits or image processing timeouts.
(* Core classification logic *)
identified = ImageIdentify[#image, All, 3];
category = categorizeWaste[First[Keys[identified]]];
severity = assessSeverity[category];Municipal admins can compute the optimal garbage truck route across all pending reports using Wolfram's Travelling Salesman Problem solver:
Scalability Note & Fallback: While Wolfram Cloud handles moderate datasets elegantly, computing Traveling Salesman tours at a massive city scale may encounter execution timeouts or rate limits. As a fallback for enterprise deployment, CleanSweep's routing architecture is designed to integrate seamlessly with dedicated routing engines like Leaflet Routing Machine or Google OR-Tools.
(* Solve TSP across all report GPS coordinates *)
geoPoints = GeoPosition[{#["lat"], #["lon"]}] & /@ parsed;
tour = FindShortestTour[geoPoints];
totalDistance = Total[Table[
GeoDistance[geoPoints[[tour[[2,i]]]], geoPoints[[tour[[2,i+1]]]]],
{i, Length[tour[[2]]] - 1}
]];┌─────────────┐ POST image ┌──────────────────────┐
│ React App │ ─────────────────→ │ Wolfram Cloud API │
│ (Home.jsx) │ ←───────────────── │ classify_waste.wl │
│ │ JSON response │ • ImageIdentify │
└─────────────┘ └──────────────────────┘
┌─────────────┐ POST coordinates ┌──────────────────────┐
│ React App │ ─────────────────→ │ Wolfram Cloud API │
│ (Admin.jsx) │ ←───────────────── │ optimize_route.wl │
│ │ Optimal ordering │ • FindShortestTour │
└─────────────┘ └──────────────────────┘
See
wolfram/README.mdfor full deployment instructions.
MIT © githubutsav
| Name | GitHub |
|---|---|
| Tushar Bajpai | @Tushar |
| Sankalp Saini | @Sankalp |
| Utsav Singh | @Utsav |