sfi-visual-intelligence/mldemos_with_webcam

ML demos. Code for detection/recognition in videos using a webcam, Python and a Raspberry pi.

Code for object and face detection with a webcam and a Raspberry Pi

The repository contains Python code for face detection, object recognition and other machine learning application using a webcam. A part of the demos are made to run on a Raspberry Pi (3 or 4).

It uses several open-source libraries where neural networks have already been pre-trained for the task at hand.

• Simple face detection and recognition uses opencv library.
• Face detection with gender, age and emotion recognition is DeepFace code : https://github.com/serengil/deepface

The code is under the MIT license.

What is inside the repository

There are Jupyter notebooks and Python scripts. The notebooks are meant to guide the user, explain the different steps, and explore the possibilities of image analysis. The Python scripts contain the code to be run directly in the command line, so that it works out of the box on the Raspberry Pi (or other platforms).

Tutorial on face detection with webcam

Here is a list of relevant webpages for Relevant websites: https://iotdesignpro.com/projects/face-recognition-door-lock-system-using-raspberry-pi https://github.com/mowshon/age-and-gender https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/auto_examples/object_detection_camera.html https://learnopencv.com/training-a-custom-object-detector-with-dlib-making-gesture-controlled-applications/ http://dlib.net/ml.html

Haar cascades, face and eyes https://towardsdatascience.com/computer-vision-detecting-objects-using-haar-cascade-classifier-4585472829a9

Raspberry pi and others: https://qengineering.eu/deep-learning-with-raspberry-pi-and-alternatives.html

Experiments on the Raspberry Pi

Enable Raspberry Pi camera

You need to run the following code so that opencv can get the image with VideoCapture:

sudo modprobe bcm2835-v4l2

See: https://stackoverflow.com/questions/29583533/videocapture-open0-wont-recognize-pi-cam/37530016#37530016

Make also sure the interface for (legacy) video is enabled in the raspberry Pi configuration sudo raspi-config -> 'Interfacing Options' -> 'Camera'

Raspberry Pi and Tensorflow

It is not straightforward to install tensorflow on Rasberry Pi. Installing with pip does not work. You can find an alternative way described on the Qengineering website: https://qengineering.eu/install-tensorflow-2.7-on-raspberry-64-os.html They have a github repo with wheels versions of Tensorflow for Raspberry Pi: https://github.com/Qengineering/TensorFlow-Raspberry-Pi_64-bit Download the version you need and run pip install downloaded_wheelfile.whl. To check your version of Linux on the Raspberri Pi, just runcat /etc/os-release, the cpu architecture is given by uname -a and Python version by python --version.

Python Scripts

Face detection

You will need the following python modules:

• Open-cv:

pip install opencv-python
pip install opencv-contrib-python

The script test_video.py is the first and simplest script you can run to test if Python with opencv can access your camera and display it. You can record your face for training using record_user_face.py after recording one or several faces, run train_face_recognition.py to train the recognition process When it has been trained, run face_detector.py.

Face recognition with Deepface

You will need the deepFace module (pip install deepface) and Tensorflow installed.

You can test the webcam stream with the python script deepFace_stream.py. On the video, you will get emotion (side), estimated age and gender (M/W) (top) of each face.