Recycling Smart

Recycling is an important practice with environmental, economic, and social benefits and should be supported both on an individual and societal level. Encouraging and teaching recycling to students in schools not only creates environmental awareness, but also enables future generations to grow up as individuals who are more sensitive to their society, responsible and able to contribute to a sustainable future. In addition to raising recycling awareness in society, technology can also be utilised to make recycling easier and more fun. Acquiring knowledge in AI-related subjects can enhance the creativity of students and boost their capacity to devise innovative solutions for the challenges they will encounter in the future.

This teaching material focuses on facilitating recycling with artificial intelligence and machine learning based-object recognition. Designing a smart recycling system increases students’ awareness and develops their artificial intelligence and robotics skills. It is based on a project carried out to encourage students of a VET school to recycle. It was prepared within the scope of the elective course Artificial Intelligence and Machine Learning and won the first prize in its category in the competition organised at Atatürk Mesleki ve Teknik Anadolu Lisesi, Ankara, Türkiye, as well as the first prize in the STEM Alliance – Lenovo Competition 2023: Turning Waste into Educational Wonder.

Recycling is a practice that has many important benefits such as protecting natural resources, saving energy, reducing the amount of waste, protecting the environment, creating employment opportunities and supporting the principle of sustainability. Thanks to recycling, the consumption of natural resources is reduced and raw materials are reused, which allows us to use natural resources in a more sustainable way, making them transferable to future generations and maintaining the ecological balance. Reducing the amount of waste eliminates the need to expand landfills or open new areas. This reduces environmental pollution and waste management problems. Finally, the recycling sector also creates employment opportunities, thus contributing to economic growth.

Video by the UN Environment Programme: #ZeroWasteDay: An opportunity to address the global waste crisis

sudo apt-get update

sudo apt-get upgrade

pip install tensorflow

pip install opencv-python

pip install numpy

pip install keras

pip install pigpio

pip install time

sudo apt-get update && sudo apt-get install python3-rpi.gpio

sudo apt-get update && sudo apt-get install python3-pigpio

sudo pigpiod

https://forums.raspberrypi.com/viewtopic.php?t=340847

CORAL USB:

echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | sudo tee /etc/apt/sources.list.d/coral-edgetpu.list

curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -

sudo apt-get update

sudo apt-get install libedgetpu1-std

sudo apt-get install libedgetpu1-max

##sudo pigpiod this command must be run again every time the raspberry pi is rebooted to activate the engines 

import os  # import libraries
import cv2
from pycoral.utils.dataset import read_label_file
from pycoral.utils.edgetpu import make_interpreter
from pycoral.adapters import common
from pycoral.adapters import classify
import pigpio
import time
import RPi.GPIO as GPIO
import numpy as np

# Define the path to the TFLite model converted for use with Edge TPU
modelPath = '/home/admin/Desktop/TF/model_edgetpu.tflite'

# Define the path to the labels file that was downloaded with the model
labelPath = '/home/admin/Desktop/TF/labels.txt'

# Define GPIO pins for servos
servo1 = 23
servo2 = 24
servo3 = 25
servo4 = 21

# Initialize PWM (Pulse Width Modulation) for servo control using pigpio library
pwm = pigpio.pi()
pwm.set_mode(servo1, pigpio.OUTPUT)
pwm.set_mode(servo2, pigpio.OUTPUT)
pwm.set_mode(servo3, pigpio.OUTPUT)
pwm.set_mode(servo4, pigpio.OUTPUT)

# Set PWM frequency for servos
pwm.set_PWM_frequency(servo1, 50)
pwm.set_PWM_frequency(servo2, 50)
pwm.set_PWM_frequency(servo3, 50)
pwm.set_PWM_frequency(servo4, 50)

# Set initial servo positions
pwm.set_servo_pulsewidth(servo1, 500)
time.sleep(1)
pwm.set_servo_pulsewidth(servo2, 500)
time.sleep(1)
pwm.set_servo_pulsewidth(servo3, 500)
time.sleep(1)
pwm.set_servo_pulsewidth(servo4, 500)
time.sleep(1)

# Function to classify an image using the provided TFLite interpreter
def classifyImage(interpreter, image):
    size = common.input_size(interpreter)
    common.set_input(interpreter, cv2.resize(image, size, fx=0, fy=0, interpolation=cv2.INTER_CUBIC))
    interpreter.invoke()
    return classify.get_classes(interpreter)

def main():
    # Load the model onto the TF Lite Interpreter
    interpreter = make_interpreter(modelPath)
    interpreter.allocate_tensors()
    labels = read_label_file(labelPath)

    # Open the video capture device
    cap = cv2.VideoCapture(0)
    last_detection_cam = 0 #"cam" stands for "glass"
    last_detection_metal = 0
    last_detection_plastik = 0 #"plastik" stands for "plastics"
    last_detection_kompost = 0 #"kompost" stands for "organic waste"

    while cap.isOpened():
        # Read a frame from the video capture device
        ret, frame = cap.read()
        if not ret:
            break

        # Flip the image horizontally
        frame = cv2.flip(frame, 1)

        # Classify and display the image
        results = classifyImage(interpreter, frame)
        cv2.imshow('frame', frame)

        # Extract information about the detected object
        detected_label = labels[results[0].id]
        confidence_score = results[0].score
        print(f'Detected Label: {detected_label}, Score: {confidence_score}')

        # Control the servos based on the detected object and confidence score
        if detected_label == 'cam' and confidence_score > 0.9:
            pwm.set_servo_pulsewidth(servo1, 1700)
            last_detection_cam = time.time()
        elif detected_label == 'metal' and confidence_score > 0.9:
            pwm.set_servo_pulsewidth(servo2, 1700)
            last_detection_metal = time.time()
        elif detected_label == 'kompost' and confidence_score > 0.9:
            pwm.set_servo_pulsewidth(servo3, 1700)
            last_detection_kompost = time.time()
        elif detected_label == 'plastik' and confidence_score > 0.9:
            pwm.set_servo_pulsewidth(servo4, 1700)
            last_detection_plastik = time.time()

        # Reset servo positions after a certain duration if no detection
        if time.time() - last_detection_cam > 4:
            pwm.set_servo_pulsewidth(servo1, 500)
        if time.time() - last_detection_metal > 4:
            pwm.set_servo_pulsewidth(servo2, 500)
        if time.time() - last_detection_kompost > 4:
            pwm.set_servo_pulsewidth(servo3, 500)
        if time.time() - last_detection_plastik > 4:
            pwm.set_servo_pulsewidth(servo4, 500)

        # Check for the 'q' key press to exit the loop
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break

    # Release the video capture device and close all OpenCV windows
    cap.release()
    cv2.destroyAllWindows()

if __name__ == '__main__':
    main()

Code examples and TensorflowLite model