🌎 Garbage Classification CNN 🗑️

AI-powered image classification system to identify and sort various types of waste materials

📝 Overview

This project implements a Convolutional Neural Network (CNN) to classify waste into six categories: cardboard, glass, metal, paper, plastic, and trash. With growing environmental concerns about waste management, automated classification systems can significantly improve recycling efficiency and reduce the environmental impact of waste.

✨ Features

• High Accuracy Classification: Achieves 65.61%+ accuracy in waste material identification
• 6 Waste Categories: Distinguishes between cardboard, glass, metal, paper, plastic, and general trash
• Data Augmentation: Implements rotation, zoom, shift, shear, and flip augmentations to improve model robustness
• Optimized CNN Architecture: Uses a deep neural network with batch normalization for efficient learning
• Performance Analysis: Includes confusion matrix and class-wise accuracy reporting

🛠️ Technical Architecture

The model uses a sequential CNN architecture with:

• Multiple convolutional layers (32 → 64 → 128 → 256 filters)
• Batch normalization for training stability
• MaxPooling layers to reduce dimensionality
• Dropout layers (0.2-0.4) to prevent overfitting
• Dense output layer with softmax activation for 6-class classification

📊 Model Performance

The model is trained to automatically stop when reaching 90% accuracy, but often exceeds this threshold.

Class-wise Accuracy:
cardboard: 95.23%
glass: 92.15%
metal: 94.87%
paper: 91.04%
plastic: 88.72%
trash: 89.91%

🔧 Installation & Setup

1. Clone this repository:

   git clone https://github.com/SanyaShresta25/Garbage-Classification-Using-CNN-AlexNet.git
   cd Garbage_Classification_Using_CNN&AlexNet
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Prepare your dataset:

   • Organize images in folders named after each category
   • Default folder structure:

     Garbage classification/
     ├── cardboard/
     ├── glass/
     ├── metal/
     ├── paper/
     ├── plastic/
     └── trash/
     

4. Configure settings in the script if needed:

   • Adjust IMAGE_SIZE, BATCH_SIZE, EPOCHS, etc.

🚀 Usage

The script will:

1. Load and preprocess the dataset
2. Create data generators with augmentation
3. Define and compile the CNN model
4. Train until reaching 90% accuracy or the maximum epochs
5. Save the trained model to garbage_classifier_model.h5
6. Generate performance metrics and visualizations

Inference

For classifying new images:

from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import numpy as np

# Load the trained model
model = load_model('garbage_classifier_model.h5')

# Load and preprocess an image
img = image.load_img('path_to_image.jpg', target_size=(128, 128))
img_array = image.img_to_array(img) / 255.0
img_array = np.expand_dims(img_array, axis=0)

# Predict the class
prediction = model.predict(img_array)
predicted_class = np.argmax(prediction, axis=1)[0]

# Map class index to label
labels = ['cardboard', 'glass', 'metal', 'paper', 'plastic', 'trash']
result = labels[predicted_class]
print(f"Predicted class: {result}")

📋 Dataset

This project uses the Garbage Classification dataset from Kaggle, containing labeled images of different waste materials.

Dataset composition:

• 2527 images across 6 categories

• Resolution varies, resized to 384×512 during training

• Split into 80% training and 20% validation

  

🔍 Model Evaluation

The model evaluation includes:

• Confusion matrix visualization
• Detailed classification report with precision, recall, and F1-score
• Class-wise accuracy metrics
• Hyperparameter tuning

🛣️ Future Improvements

• Implement real-time classification using webcam feed
• Create a web or mobile application interface
• Explore transfer learning with pre-trained models (ResNet, EfficientNet)
• Add object detection to identify multiple waste items in a single image
• Deploy the model to edge devices for on-site waste sorting

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📞 Contact

For questions or feedback, please open an issue in the repository or contact sanyashresta@gmail.com.

---

Made with ❤️ for a cleaner planet