Face Detection with Open CV | Unsupervised learning in Python

Motive

     

        Machine Learning is not that easy to lean in python and YouTube tutorials cannot rely heavily so I taught of making a model and explain to you guys how machine learning works. We will not be making the rules for the model instead we will take a pre-trained model for our use. Training a model takes time and we need a lot of data so we will take the easier path for now. 

The Project

        The only library that we need is OpenCV, if you have not installed it then pip install it in your terminal, pip install opencv-python. You can learn more here.

Now open your preferred ide and import opencv, while coding we use cv2 for opencv:

# import opencv import cv2

        Now as I told above we will use a pre-trained model for this use so download it from here. Now let's mention it in the code and ensure that the XML file is in the same directory.

import cv2 faceDetection = cv2.CascadeClassifier('faceDetection.xml')

        We will be using the webcam to detect the faces so let's mention that also with the help of open cv it is easy to take frames from a video and use it for detection. 

import cv2 faceDetection = cv2.CascadeClassifier('faceDetection.xml') camera = cv2.VideoCapture(0)

        If you want to use a file instead of the use the file name in inverted commas instead of the 0. Like this cv2.VideoCapture('filename.mp4').

Since it is a video we have to take its frames for detection for that we need a while loop, In which we take frames from the video and then convert it into grayscale.

while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

        Here we take the frames for the video in the variable image and the cover the image to greyscale that greyscale image is stored in the variable called grayscale.

Now let's detect the faces:

while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) faces = faceDetection.detectMultiScale(grayscale, 1.2, 4)  

        The three variables are, grayscale which is the image,1.2 which is the scale factor and 4 is the minimum neighbours. If any issues change these values.

 Now, let's draw a black rectangle for the face:

while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) faces = faceDetection.detectMultiScale(grayscale, 1.2, 4) for (a, b, w, h) in faces: cv2.rectangle(img, (a, b), (a+w, b+h), (0, 0, 0), 2)

 Now show the image.

while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) faces = faceDetection.detectMultiScale(grayscale, 1.2, 4) for (a, b, w, h) in faces: cv2.rectangle(image, (a, b), (a+w, b+h), (0, 0, 0), 2) cv2.imshow('image', image)

Now if see key is pressed it loop will break and the code ends.

while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) faces = faceDetection.detectMultiScale(grayscale, 1.2, 4) for (a, b, w, h) in faces: cv2.rectangle(image, (a, b), (a+w, b+h), (0, 0, 0), 2) cv2.imshow('image', image) esc = cv2.waitKey(30) & 0xff if esc==27: break camera.release()

And that's it and run the code and test it your self.

The Full Code

import cv2 faceDetection = cv2.CascadeClassifier('faceDetection.xml') camera = cv2.VideoCapture(0) while True: _, image = camera.read() grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) faces = faceDetection.detectMultiScale(grayscale, 1.2, 4) for (a, b, w, h) in faces: cv2.rectangle(image, (a, b), (a+w, b+h), (0, 0, 0), 2) cv2.imshow('image', image) esc = cv2.waitKey(30) & 0xff if esc==27: break camera.release()

      If you have any doubts or errors in the program feel free to comment down below and I will respond as early as possible. And if I have made any mistake please correct me in the comment box below. Don't forget to follow us for such python related content.