🌿 Sutham — Smart Waste Intelligence for Bharat

सुथम · சுத்தம் · Clean

AI-powered waste detection, hazard intelligence, and route optimization platform
built for Coimbatore Municipal Corporation — and every Indian city after it.

🏆 One photo. Ten seconds. A dispatch-ready cleanup ticket.

93 Active Sites	3,194 kg Detected	96% AI Confidence	< 10s Per Ticket
Live · Coimbatore	Total Load	Nova Pro Vision	End-to-End

🚨 The Problem

India generates 150,000 tonnes of municipal solid waste every single day.
The waste management market is worth $8 billion and growing at 6% annually — yet the system running it is WhatsApp messages, phone calls, and paper registers.

Pain Point	Current Reality	Sutham's Fix
No real-time visibility	Supervisors fly blind	Live AI-mapped hotspots
Unstructured complaints	WhatsApp photos, calls	Structured JSON dispatch tickets
No priority intelligence	Every complaint equal	P0 / P1 / P2 AI triage
Sensitive zones ignored	Schools, hospitals at risk	Automatic proximity alerts
Manual routing	Driver decides the path	AI-optimized multi-stop routes
Hazards undetected	Dengue, fire, animals missed	Layered hazard overlays

💡 Our Solution

Sutham is a two-layer AI waste intelligence platform:

 CITIZEN / FIELD WORKER                    MUNICIPAL SUPERVISOR
 ─────────────────────                    ────────────────────────
  📸 Snap a photo                          🗺️  EcoRoute Command Dashboard
       ↓                                        ↓
  🤖 Nova Pro Vision                        📊 93 live sites mapped
  analyzes waste type,                      🔴 40 P0 Critical flagged
  volume, weight,                           ⚠️  45 hazard sites overlaid
  recyclables, priority                     🏫 33 near sensitive zones
       ↓                                        ↓
  📍 GPS auto-extracted                     🚛 Optimized route generated
  AWS Location geocoding                    in <500ms for 100 stops
  Sensitive zone check                      764 kg load dispatched
       ↓
  🎫 Ticket ID generated
  Priority + Vehicle + Action
  → Ready for EcoRoute in seconds

🖥️ Platform Overview

1 · User Interface — Sutham Next (Mobile-First Reporting)

The citizen and field-worker facing app. Powered by Next.js 16 with AWS Bedrock Nova Pro.

Key Capabilities:

📸 Single photo upload → full AI waste analysis in under 10 seconds
🛰️ Automatic GPS from EXIF data; device GPS fallback
🗺️ AWS Location Service reverse-geocoding to exact street address
🏥 Real-time sensitive zone proximity check (hospitals, schools, PHCs, water bodies)
♻️ Recyclable material detection (plastic, wood, metal, organic)
📋 Structured JSON ticket output: type · weight · volume · priority · vehicle · reasoning
🔒 Confidence score + image quality assessment per upload

Sample AI Output:

{
  "waste_detected": true,
  "waste_type": "MIXED",
  "volume_level": "HIGH",
  "estimated_weight_kg": 50,
  "area_covered_sqm": 5,
  "priority": "P1",
  "action": "DISPATCH_NOW",
  "vehicle_type": "PICKUP",
  "high_value_recyclables_present": true,
  "recyclable_types": ["PLASTIC", "WOOD"],
  "confidence_score": 0.95,
  "near_sensitive_zone": false,
  "reasoning": "Significant mixed waste including plastic bottles and wooden crates spread over 5 sqm. High-value recyclables present. Pickup dispatch recommended."
}

2 · Admin Dashboard — EcoRoute (Command & Dispatch)

The municipal supervisor's command center. Built with React 18 + Vite + React-Leaflet.

Key Capabilities:

🗺️ Interactive live map with 1000+ stop support at smooth 60fps
🏙️ Zone-based operations: North · South · East · West · Central · Industrial
🔴 Priority queue: P0 Critical → P1 High → P2 Routine with visual differentiation
🛣️ AI route optimizer: Greedy Nearest Neighbor TSP, <500ms for 100 stops
📍 Custom GPS pin + 10 predefined depot landmarks + draggable map pins
⚠️ Hazard layer overlays: Dengue/Stagnant Water · Air Quality/Smoke · Animal Scavenging
🏥 Sensitive zone alerts with distance-to-school/hospital display
📊 Live shift summary: Dispatch Now count · Route stops · Total load (kg)
☁️ AWS S3 dataset integration with localStorage cache fallback

🏗️ Architecture

sutham/
├── admin-ui/                          # React + Vite Command Dashboard
│   ├── src/
│   │   ├── components/
│   │   │   ├── Map.jsx               # Leaflet map with all overlays
│   │   │   ├── SidePanel.jsx         # Zone filter + route controls
│   │   │   ├── StopDetails.jsx       # Right panel: site info + AI reasoning
│   │   │   └── ShiftSummary.jsx      # Live KPI bar
│   │   └── utils/
│   │       ├── haversine.js          # GPS distance (Haversine formula)
│   │       ├── routeOptimizer.js     # Nearest Neighbor TSP
│   │       ├── zones.js              # Zone assignment + colors
│   │       └── wards.js              # CCMC ward boundary lookup
│   ├── Dockerfile                    # Nginx production container
│   └── .env.example
│
└── user-ui/
    └── WasteManagement/
        └── User-interface/            # Next.js 16 Mobile App
            ├── src/
            │   ├── app/
            │   │   └── api/
            │   │       └── analyze/  # Nova Pro + Location Service route
            │   └── components/
            │       ├── UploadZone    # Drag & drop + camera capture
            │       ├── TicketCard    # AI result display
            │       └── MapPreview    # GPS location preview
            ├── docs/
            │   └── DEPLOYMENT_GUIDANCE.md
            └── .env.example

🔧 Tech Stack

Admin Dashboard

Layer	Technology
Frontend	React 18, Vite 7
Mapping	React-Leaflet 4, Leaflet 1.9, CartoDB tiles
Styling	Tailwind CSS 3
Icons	Lucide React
Routing Algorithm	Greedy Nearest Neighbor TSP (Haversine)
Cloud	AWS S3, AWS SDK v3
Containerization	Docker + Nginx

User Interface

Layer	Technology
Framework	Next.js 16 (App Router)
Runtime	React 19
Language	TypeScript (full type safety)
AI Vision	AWS Bedrock — Amazon Nova Pro v1
Geocoding	AWS Location Service (Places API v2)
GPS	EXIF extraction → device fallback
Styling	Tailwind CSS 4
Deployment	AWS App Runner / Amplify

🗺️ Zone Management

Coimbatore divided into 6 operational zones with dedicated vehicle fleets:

Zone	Color	Vehicle	Coverage Rule
🔴 NORTH	Red	Compactor	lat ≥ 11.060
🟣 SOUTH	Purple	E-Rickshaw	lat ≤ 10.980
🔵 EAST	Blue	Tipper	lng ≥ 77.050
🩵 WEST	Teal	Mini Truck	lng ≤ 76.940
🟠 CENTRAL	Orange	Compactor x2	Central business district
⚫ INDUSTRIAL	Grey	Heavy Truck	Industrial corridors

🤖 AI Analysis Pipeline

Image Upload (JPG/PNG/TIFF/HEIC up to 200MB)
        ↓
EXIF GPS Extraction (Pillow)
   → if no EXIF: Device GPS fallback
        ↓
Image → PNG conversion → AWS Bedrock (Nova Pro v1:0)
   Tool-driven agent: forces single waste_report() call
        ↓
Structured JSON output:
   waste_type · volume_level · estimated_weight_kg
   area_covered_sqm · priority · action · vehicle_type
   stagnant_water · smoke · animal · recyclable_types
   confidence_score · image_quality · reasoning
        ↓
AWS Location Service (reverse-geocode + search-nearby)
   → street address · suburb · city · state
   → nearby landmarks (hospitals, schools, stations)
        ↓
Sensitive Zone Check (Haversine · 47 verified zones)
   → hospitals · PHCs · schools · colleges · water bodies
   → if within 50m: auto-escalate to P0
        ↓
Final Ticket JSON → S3 storage → EcoRoute Dashboard

📦 Dataset Schema

Every waste collection point follows this normalized schema:

{
  "ticket_id":                   "uuid-string",
  "created_at":                  "ISO-8601 timestamp",
  "area_name":                   "4th Street, Coimbatore, Tamil Nadu",
  "lat":                         11.041895,
  "lng":                         77.044629,
  "priority":                    "P0 | P1 | P2",
  "action":                      "DISPATCH_NOW | ADD_TO_ROUTE",
  "waste_type":                  "MIXED | ORGANIC | PLASTIC | METAL | ...",
  "volume_level":                "CRITICAL | HIGH | MEDIUM | LOW",
  "estimated_weight_kg":         50,
  "area_covered_sqm":            5,
  "vehicle_type":                "COMPACTOR | PICKUP | TIPPER | ...",
  "stagnant_water_detected":     false,
  "smoke_detected":              false,
  "animal_presence":             false,
  "near_sensitive_zone":         true,
  "nearest_zone": {
    "name":                      "St Michael's Academy",
    "zone_type":                 "SCHOOL",
    "distance_m":                26.5
  },
  "high_value_recyclables_present": true,
  "recyclable_types":            ["PLASTIC", "WOOD"],
  "image_file":                  "images/uuid.png",
  "confidence_score":            0.95,
  "image_quality":               "CLEAR",
  "gps_source":                  "exif | device",
  "gps_accuracy_m":              20,
  "wall_time_seconds":           6,
  "reasoning":                   "Natural language AI explanation"
}

🚀 Quick Start

Prerequisites

Node.js 18+ (Admin) / 20+ (User Interface)
AWS Account with: S3 · Bedrock (Nova Pro) · Location Service · IAM credentials

Admin Dashboard

git clone https://github.com/your-org/sutham.git
cd sutham/admin-ui

npm install
cp .env.example .env
# Add your AWS credentials to .env

npm run dev
# → http://localhost:5173

User Interface

cd sutham/user-ui/WasteManagement/User-interface

npm install
cp .env.example .env.local
# Add your AWS credentials to .env.local

npm run dev
# → http://localhost:3000

🔐 Environment Variables

Admin Dashboard — `.env`

AWS_REGION=us-east-1
AWS_ACCESS_KEY_ID=your_access_key
AWS_SECRET_ACCESS_KEY=your_secret_key
S3_BUCKET=your-sutham-bucket
S3_KEY=input/dataset_synthetic_100.json

User Interface — `.env.local`

AWS_ACCESS_KEY_ID=your_access_key
AWS_SECRET_ACCESS_KEY=your_secret_key
AWS_DEFAULT_REGION=us-east-1
AWS_LOCATION_API_KEY=your_location_api_key
AWS_LOCATION_REGION=us-east-1

⚠️ Never commit credentials to source control. Use AWS IAM roles in production.

📦 Deployment

Admin Dashboard

Method	Command	Notes
Dev server	`npm run dev`	Port 5173
Production build	`npm run build`	Static output
Docker	`docker build -t sutham-admin .`	Nginx container
AWS ECS	See `Dockerfile`	Recommended for prod

User Interface

Method	Notes
AWS Amplify	Recommended · managed Next.js · auto CI/CD
AWS App Runner	Containerized · currently live
AWS ECS Fargate	Full control · production scale

See DEPLOYMENT_GUIDANCE.md for complete step-by-step instructions.

📈 Performance

Metric	Target	Achieved
AI ticket generation	< 10s	✅ ~6s avg
Route optimization (100 stops)	< 500ms	✅
Map rendering (500+ markers)	60fps smooth	✅
Admin dashboard stop capacity	1000+	✅
GPS accuracy (EXIF)	± 5m	✅
GPS accuracy (device fallback)	± 20–35m	✅

🎯 Use Cases

Municipal Operations — Real-time fleet dispatch and optimized multi-zone routing
Citizen Reporting — One-photo waste reporting with zero form-filling
Health & Hazard Monitoring — Dengue, smoke, and animal scavenging alerts
Sensitive Zone Compliance — Automatic escalation near hospitals, schools, PHCs
Circular Economy — Recyclable hotspot intelligence for buyback operators
Shift Planning — Data-driven vehicle allocation and workload balancing

🤝 Acknowledgments

Coimbatore Municipal Corporation — for the operational context and zone data
Amazon Web Services — Nova Pro vision, Location Service, S3, App Runner
OpenStreetMap contributors — base map data
CartoDB — map tile hosting

Built with ❤️ for AI for Bharat Sustainability Hackathon 2026

Sutham — सुथम — சுத்தம் — Clean

Name		Name	Last commit message	Last commit date
Latest commit History 33 Commits
User-interface		User-interface
admin-page		admin-page
README.md		README.md

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

🌿 Sutham — Smart Waste Intelligence for Bharat

🏆 One photo. Ten seconds. A dispatch-ready cleanup ticket.

📖 Table of Contents

🚨 The Problem

💡 Our Solution

🖥️ Platform Overview

1 · User Interface — Sutham Next (Mobile-First Reporting)

2 · Admin Dashboard — EcoRoute (Command & Dispatch)

🏗️ Architecture

🔧 Tech Stack

Admin Dashboard

User Interface

🗺️ Zone Management

🤖 AI Analysis Pipeline

📦 Dataset Schema

🚀 Quick Start

Prerequisites

Admin Dashboard

User Interface

🔐 Environment Variables

Admin Dashboard — `.env`

User Interface — `.env.local`

📦 Deployment

Admin Dashboard

User Interface

📈 Performance

🎯 Use Cases

🤝 Acknowledgments

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

🌿 Sutham — Smart Waste Intelligence for Bharat

🏆 One photo. Ten seconds. A dispatch-ready cleanup ticket.

📖 Table of Contents

🚨 The Problem

💡 Our Solution

🖥️ Platform Overview

1 · User Interface — Sutham Next (Mobile-First Reporting)

2 · Admin Dashboard — EcoRoute (Command & Dispatch)

🏗️ Architecture

🔧 Tech Stack

Admin Dashboard

User Interface

🗺️ Zone Management

🤖 AI Analysis Pipeline

📦 Dataset Schema

🚀 Quick Start

Prerequisites

Admin Dashboard

User Interface

🔐 Environment Variables

Admin Dashboard — .env

User Interface — .env.local

📦 Deployment

Admin Dashboard

User Interface

📈 Performance

🎯 Use Cases

🤝 Acknowledgments

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Admin Dashboard — `.env`

User Interface — `.env.local`

Packages