How to Implement Real-Time Object Detection on Raspberry Pi 4 with TensorFlow Lite

Implementing Real-Time Object Detection on Raspberry Pi 4 Using TensorFlow Lite and USB Webcam Stream

In this tutorial, you’ll learn how to implement real-time object detection on a Raspberry Pi 4 using TensorFlow Lite with a USB webcam. This setup is ideal for projects involving computer vision, robotics, and IoT applications.

Prerequisites

• Raspberry Pi 4 with Raspbian OS installed
• USB webcam compatible with Raspberry Pi
• Internet connection for installation
• Basic knowledge of Python programming
• Access to a terminal or SSH client

Parts/Tools

• Raspberry Pi 4
• USB webcam
• Micro SD card (at least 16GB recommended)
• Power supply for Raspberry Pi
• Keyboard and monitor (or SSH access)
• Python 3 installed
• TensorFlow Lite and other necessary libraries

Steps

1. Set Up Your Raspberry Pi
   1. Ensure your Raspberry Pi is powered on and connected to the internet.
   2. Open a terminal window.
   3. Update your package list:

   sudo apt update

   4. Upgrade installed packages:

   sudo apt upgrade

2. Install Required Libraries
   1. Install OpenCV for handling video streams:

   sudo apt install python3-opencv

   2. Install TensorFlow Lite dependencies:

   pip3 install tensorflow tensorflow-hub

3. Connect the USB Webcam
   1. Plug the USB webcam into one of the USB ports on the Raspberry Pi.
   2. Verify the webcam is recognized by the system:

   ls /dev/video*

4. Download TensorFlow Lite Model
   1. Choose a pre-trained model, such as the SSD MobileNet v2 model.
   2. Download the model files:

   wget https://storage.googleapis.com/download.tensorflow.org/models/tflite/gpu/lite-models/ssd_mobilenet_v2/1/default/1.tflite

5. Create the Object Detection Script
   1. Create a new Python file for your script:

   nano object_detection.py

   2. Copy and paste the following code snippet into the file:

   import cv2
   import numpy as np
   import tensorflow as tf
   
   # Load the TFLite model and allocate tensors.
   interpreter = tf.lite.Interpreter(model_path="1.tflite")
   interpreter.allocate_tensors()
   
   # Get input and output tensors.
   input_details = interpreter.get_input_details()
   output_details = interpreter.get_output_details()
   
   # Open webcam stream
   cap = cv2.VideoCapture(0)
   
   while True:
       ret, frame = cap.read()
       if not ret:
           break
       
       # Preprocess the image
       input_data = cv2.resize(frame, (300, 300))
       input_data = np.expand_dims(input_data, axis=0)
       
       # Run inference
       interpreter.set_tensor(input_details[0]['index'], input_data)
       interpreter.invoke()
       
       # Process the output
       boxes = interpreter.get_tensor(output_details[0]['index'])
       classes = interpreter.get_tensor(output_details[1]['index'])
       scores = interpreter.get_tensor(output_details[2]['index'])
       
       # Display results
       for i in range(len(scores)):
           if scores[i] > 0.5:
               box = boxes[i]
               cv2.rectangle(frame, (int(box[1]*frame.shape[1]), int(box[0]*frame.shape[0])), 
                                   (int(box[3]*frame.shape[1]), int(box[2]*frame.shape[0])), (0, 255, 0), 2)
       
       cv2.imshow('Object Detection', frame)
       if cv2.waitKey(1) & 0xFF == ord('q'):
           break
   
   cap.release()
   cv2.destroyAllWindows()

   3. Save the file and exit the editor.
6. Run the Object Detection Script
   1. Execute the script in the terminal:

   python3 object_detection.py

   2. Make sure the webcam is streaming video and the detected objects appear in the window.

Troubleshooting

• Webcam Not Detected: Ensure the webcam is properly connected. Check the output of ls /dev/video* to confirm.
• Model Not Found: Verify the model file path is correct and that the file exists in your working directory.
• Low Frame Rate: Reduce the resolution of the video feed or optimize the model for better performance.
• Errors in Script Execution: Check for syntax errors or missing libraries. Ensure Python 3 is installed and the necessary packages are present.

Conclusion

You’ve successfully set up real-time object detection on your Raspberry Pi 4 using TensorFlow Lite and a USB webcam. This project can be expanded with additional features like saving detected objects or integrating with other IoT devices. Happy coding!