GEOSMART – Flask Application

A waste-intelligence command centre migrated from Vite/React to Python Flask. The UI, styling, and functionality remain identical to the original React build.

Project Structure

flask_app/
├── app.py                 # Flask backend with all routes & API endpoints
├── requirements.txt       # Python dependencies
├── templates/
│   ├── base.html          # Shared layout (Tailwind CDN, Lucide icons, toast)
│   ├── splash.html        # Splash screen → auto-redirects to login
│   ├── login.html         # Login form
│   ├── signup.html        # Signup form
│   ├── dashboard.html     # Main dashboard with metrics & navigation
│   ├── index_page.html    # Landing / marketing page
│   ├── analyze.html       # Waste stream analyzer with pie chart
│   ├── complaint.html     # Complaint lodge & status board
│   ├── municipal.html     # Municipal officer directory & schedules
│   ├── profile.html       # User profile editor
│   └── 404.html           # Not-found page
└── static/
    ├── css/
    │   └── styles.css     # Design tokens, component styles, animations
    └── js/
        ├── analyze.js     # Client-side waste computation + Chart.js
        ├── complaint.js   # Complaint CRUD via Flask API
        ├── municipal.js   # Officer search / filter
        └── profile.js     # Profile edit & save via Flask API

Prerequisites

• Python 3.9 or later

Setup & Run

# 1. Navigate to the project directory
cd flask_app

# 2. Create a virtual environment (recommended)
python -m venv venv

# 3. Activate the virtual environment
#    Windows:
venv\Scripts\activate
#    macOS / Linux:
source venv/bin/activate

# 4. Install dependencies
pip install -r requirements.txt

# 5. Run the application
python app.py

The app starts at http://127.0.0.1:5000.

Authentication Modes

The app supports two login paths:

• User Login: /login
• Admin Login: /admin/login

Admin credentials are environment-driven (defaults below):

• ADMIN_EMAIL=admin@geosmart.local
• ADMIN_PASSWORD=admin123

Set custom credentials before running if needed:

# Windows CMD
set ADMIN_EMAIL=you@domain.com
set ADMIN_PASSWORD=yourStrongPassword
python app.py

Model Training (Admin)

Admins can train the classification model from the dashboard or Analyze page.

1. Login via /admin/login
2. Open dashboard training panel
3. Enter dataset YAML path (Ultralytics format)
4. Set epochs and image size
5. Click Start model training

On the Analyze page admin panel, you can also:

• Validate a local data.yaml path before training
• Upload a dataset .zip archive; the app auto-detects data.yaml and fills the training path

Additional model APIs:

• POST /api/model/validate-dataset (admin)
• POST /api/model/upload-dataset-zip (admin)
• POST /api/analyze/cnn (authenticated image prediction)

When training finishes, best weights are copied to:

• models/waste_model.pt

This model is then used for image verification in complaint uploads.

Routes

URL	Method	Description
/	GET	Splash screen (redirects to login)
/login	GET / POST	Login page
/signup	GET / POST	Signup page
/dashboard	GET	Main dashboard
/index	GET	Landing page
/analyze	GET	Waste analysis tool
/complaint	GET	Complaint desk
/municipal	GET	Municipal response hub
/profile	GET	User profile
/logout	GET	Clear session & redirect

API Endpoints

URL	Method	Description
/api/complaints	GET	List all complaints
/api/complaints	POST	Create a complaint
/api/complaints/<id>/status	PATCH	Update complaint status
/api/profile	GET	Get profile data
/api/profile	PUT	Update profile data

Technology Stack

• Backend: Python Flask
• Styling: Tailwind CSS (CDN) with custom HSL design tokens
• Icons: Lucide (CDN)
• Charts: Chart.js (CDN)
• Templating: Jinja2

Paper Alignment (GeoSmart)

This codebase implements the core workflow described in your paper:

• Citizen reporting with image + GPS: Complaint form supports evidence upload and location capture.
• AI-based verification: POST /api/analyze/cnn verifies uploaded images and returns predicted waste type + confidence.
• Hotspot detection (GIS-aware): Hotspots are generated from complaint geolocation with severity scoring and queue ranking.
• Route optimization (A):* Municipal hub calls GET /api/routes/optimize to generate shortest dispatch plans from nearest depot.
• Live dashboards: Dashboard, complaint desk, and municipal hub auto-refresh metrics/hotspots periodically.
• Lifecycle tracking: Complaint status flow is aligned to: Pending -> Assigned -> In Progress -> Resolved.
• Admin model training: Dataset validation, zip upload, and model retraining endpoints are available from dashboard/analyze pages.

If you want to match the paper further, the next optional additions are:

• IoT bin ingestion API for sensor streams.
• True multi-stop VRP optimizer beyond single-hotspot route selection.
• Traffic-aware ETA integration using an external roads/traffic API.