Smart-Waste-Segregation-System

An AI-powered web application that automatically classifies waste materials into 12 categories and provides tailored recycling guidance to promote sustainable waste management practices.

Features

• Image Classification: Upload images or use live camera to identify waste
• AI-Powered: Deep learning model based on MobileNetV2 architecture
• Recycling Guidance: Detailed recycling methods for each waste category
• Location-Based: Find recycling centers in your state
• QR Integration: Quick access to recycling centers via QR codes
• Camera Support: Real-time waste classification using device camera
• Confidence Scores: Transparent AI predictions with probability percentages
• Web-Based: Accessible from any device with a browser

Waste Categories Classified

1. Battery
2. Biological/Organic Waste
3. Brown Glass
4. Cardboard
5. Clothes/Textiles
6. Green Glass
7. Metal
8. Paper
9. Plastic
10. Shoes
11. Trash (Non-recyclable)
12. White Glass

Tech Stack

Frontend

• Streamlit - Web application framework
• OpenCV - Image processing and computer vision
• QRCode - QR code generation for recycling centers

Backend & AI/ML

• TensorFlow - Machine learning framework
• Keras - Deep learning API
• MobileNetV2 - Pre-trained CNN model for transfer learning
• NumPy - Numerical computing and array processing

Deployment

• Streamlit Sharing - Cloud deployment platform
• GitHub - Version control and code hosting

Installation

Prerequisites

• Python 3.8 or higher
• pip (Python package manager)

Steps

1. Clone the repository

   git clone https://github.com/your-username/Smart-Waste-Segregation-System.git
   cd Smart-Waste-Segregation-System
   

Usage

Choose Input Method: Select between image upload or camera capture

Provide Waste Image: Upload a clear image or use your camera to capture waste

Get Classification: AI model predicts the waste category with confidence score

View Recycling Guidance: Get detailed recycling instructions for the identified waste

Find Recycling Centers: Locate nearby recycling facilities using the interactive map

Model Training

The classification model was trained using transfer learning with MobileNetV2 as the base model:

Dataset: Garbage Classification Dataset from Kaggle

Classes: 12 waste categories

Input Size: 224×224 pixels

Preprocessing: Image normalization (0-1 scaling)

Augmentation: Rotation, flipping, zooming, and shifting

Training: Fine-tuning with custom classification layers

Validation: 80-20 split with early stopping