/pose-detect-raspberry-pi

Primary LanguagePythonMIT LicenseMIT

Pose Detection System

This repository contains code for a real-time pose detection system that uses PoseNet, a deep learning model, to detect human poses in images or video streams. The system has two main components:

  1. Web-Based Pose Detection: A Flask-based web application that streams live video from a camera and displays it in a web browser. It also runs pose detection on the video feed and renders the detected poses on the web page in real-time. The pose detection process is performed using the PoseNet model.

  2. Local Pose Detection: A command-line script that captures video from a camera and displays it using OpenCV. It also performs pose detection on the video feed, displaying the detected poses in a graphical window.

Hardware and Software Requirements

To run this project, you will need the following hardware and software:

Hardware:

  • Raspberry Pi 4 (2GB or higher recommended)
  • Coral USB Accelerator
  • A compatible USB-C power supply for the Raspberry Pi
  • microSD card (16GB or higher)
  • Raspberry Pi Camera Module v3

Software:

  • Raspberry Pi OS (Raspbian version: 11 (bullseye) - 64)

Real-Time Pose Detection on Raspberry Pi 4

This code is optimized for real-time execution on a Raspberry Pi 4. The Raspberry Pi 4, with its improved processing power and memory, is capable of running the pose detection model with acceptable frame rates for real-time applications.

The use of the Coral USB Accelerator helps accelerate the pose detection process, making it suitable for real-time performance. The code has been fine-tuned to take advantage of the hardware capabilities of the Raspberry Pi 4 and the Coral USB Accelerator, ensuring smooth and efficient pose detection.

Installation

To run this code, you need to set up your Python environment and install the necessary dependencies. You can do this by creating a virtual environment and installing the required packages listed in the requirements.txt file.

pip install -r requirements.txt

Usage

Web-Based Pose Detection

  1. To start the web-based pose detection application, run the stream_on_web.py script. This will start a Flask web server that serves the video stream and processes pose detection.
python stream_on_web.py
  1. Open a web browser and navigate to http://localhost:5000 to access the live video stream with pose detection. Press 'q' to quit the application.

Local Pose Detection

  1. To run the local pose detection script, execute the stream_on_cv.py script.
python stream_on_cv.py
  1. This will open a window displaying the live video feed from your camera with real-time pose detection. Press 'q' to quit the application.

Model and Configuration

The PoseNet model used for pose detection is loaded from the "posenet_resnet_50_416_288_16_quant_edgetpu_decoder.tflite" file, which should be present in the "PoseNet/model" directory.

You can adjust various parameters like the frame size, accuracy threshold, and video recording settings in the detect_pose function in the posenet/pose.py script.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

The PoseNet model used in this project is based on Google's MediaPipe Pose model. The code has been adapted and modified for this project. For more details on the original model, visit MediaPipe.

Author

This project was created by [Your Name]. Feel free to reach out for questions or contributions.

Enjoy exploring and using the real-time pose detection system on your Raspberry Pi 4!