Waste Classification System with Real-Time Camera and Servo Control

This project integrates a machine learning-based waste classification system with real-time camera processing and Arduino-controlled servo motors. It classifies waste into biodegradable and non-biodegradable categories and controls hardware actions based on the classification

Features

• Machine Learning Model:

  • Built using TensorFlow/Keras with a CNN model for classifying waste.
  • Classifies waste into categories like plastic, paper, metal, cardboard, glass, compost, and trash
  • Maps waste categories into biodegradable and non_biodegradable

• Real-Time Camera Integration:

  • Captures video feed from a computer or external camera using OpenCV.
  • Preprocesses frames in real-time for model predictions

• Arduino Servo Control:

  • Communicates with an Arduino via serial to control servos for waste sorting.
  • Sends commands like biodegradable or non_biodegradable to Arduino based on predictions

• Dynamic Decision Logic:

  • Implements a confidence threshold to ensure accurate predictions.
  • Processes predictions over a time interval for better reliability

---

Requirements

Software

• Python 3.8+
• TensorFlow/Keras
• OpenCV
• NumPy
• pySerial
• A pre-trained TensorFlow model (waste_classification_model.h5).

Hardware

• A camera (webcam or external camera).

• Arduino with a servo motor attached.

• A system capable of running Python scripts.

  

---

Setup

1. Clone the Repository:

   git clone https://github.com/your-repository/waste-classification.git
   cd waste-classification

2. Install Dependencies:

   pip install tensorflow opencv-python numpy pyserial

3. Prepare the Model:

   • Place the pre-trained TensorFlow model (waste_classification_model.h5) in the project directory.
   • If you don't have a model, train one using the dataset and code provided.

4. Connect Arduino:

   • Upload a compatible Arduino sketch for servo control to the Arduino board.
   • Connect the board to your computer via USB.

5. Run the Script:

   python waste_classification.py

---

Usage

1. Camera Feed:

   • The script captures real-time video frames from the connected camera.
   • Displays the frame with prediction results overlayed.

2. Waste Classification:

   • The ML model predicts the type of waste with confidence levels.
   • The predictions are shown on the video feed.

3. Servo Motor Control:

   • Based on classification, the system sends signals to the Arduino to trigger servo actions.

4. Stop the Program:

   • Press q to quit the real-time video feed and end the program safely.

---

Directory Structure

waste-classification/
├── waste_classification.py    # Main script for the project
├── waste_classification_model.h5 # Pre-trained TensorFlow model
├── README.md                  # Project documentation
├── requirements.txt           # Python dependencies
└── arduino_sketch/            # Arduino code for servo control

---

Troubleshooting

• Camera Not Detected:
  • Ensure the camera is properly connected and recognized by your system.
• Model File Not Found:
  • Verify the waste_classification_model.h5 is in the project directory.
• Arduino Connection Fails:
  • Check the COM port and baud rate in the script.
• Prediction Accuracy Low:
  • Fine-tune the model with a larger dataset or adjust the confidence threshold.

---

DEMO

Prototype.Demo.mp4

---

Future Enhancements

• Integrate a mobile app for remote monitoring.
• Use advanced models like VGG16 or ResNet for improved accuracy.
• Add sound alerts for classification categories.

---

Acknowledgments

• TensorFlow for providing robust machine learning frameworks.
• OpenCV for real-time image processing capabilities.
• Arduino for enabling seamless hardware integration

---