[Solved] Help needed for face detection -deep learning



  • I have a code for detecting faces but now i need to count no:of faces .link text

    # OpenCV Python program to detect cars in video frame
    # import libraries of python OpenCV 
    import cv2
    #import numpy as np
    # capture frames from a video
    cap = cv2.VideoCapture('video.avi')
     
    # Trained XML classifiers describes some features of some object we want to detect
    car_cascade = cv2.CascadeClassifier('cars.xml')
     
    # loop runs if capturing has been initialized.
    while True:
        # reads frames from a video
        ret, frames = cap.read()
         
        # convert to gray scale of each frames
        gray = cv2.cvtColor(frames, cv2.COLOR_BGR2GRAY)
         
        #color fill white
        frames.fill(255)
        # or img[:] = 255
        # Detects cars of different sizes in the input image
        cars = car_cascade.detectMultiScale(gray, 1.1, 1)
         
        # To draw a rectangle in each cars
        for (x,y,w,h) in cars:
            cv2.rectangle(frames,(x,y),(x+w,y+h),(0,0,0),-1)
     
       # Display frames in a window 
        
        cv2.imshow('video2', frames)
         
        # Wait for Esc key to stop
        if cv2.waitKey(33) == 27:
            break
     
    # De-allocate any associated memory usage
    cv2.destroyAllWindows()
    


  • Hi @Nandu, You can increment a variable each time when detect faces, is that help?



  • @salmanfaris yeah that's what i want.But i am not able to understand where i should place my variable in the above mentioned code.



  • Follow These steps

    1. Create a virtual enviroment and activate virtial environment
    python -m venv venv
    

    Activate venv for windows using following command:

    .\venv\Scripts\activate
    

    For Ubuntu:

    source venv/bin/activate
    
    1. Install necessary packages on venv
    pip install opencv-python
    
    pip install imutils
    
    1. Create Folder structure as shown below in your workspace
    TestPrograms  
    |
    ├─ cascades
    │  └─ haarcascade_frontalface_default.xml
    ├─ detect_faces.py
    ├─ images
    │  └─ obama.jpg
    ├─ utilities
    │  └─ facedetector.py
    
    
    1. Program for utililities/facedetector.py given below:
    import cv2
    class FaceDetector:
        def __init__(self, face_cascade_path):
            # Load the face detector
            self.face_cascade = cv2.CascadeClassifier(face_cascade_path)
    
        def detect(self, image, scale_factor=1.2, min_neighbors=3):
            # Detect faces in the image
            boxes = self.face_cascade.detectMultiScale(image, scale_factor, min_neighbors, flags=cv2.CASCADE_SCALE_IMAGE, minSize=(30,30))
    
            # Return the bounding boxes
            return boxes
    
    1. program on detect_faces.py
    from utilities.facedetector import FaceDetector
    import imutils
    import cv2
    
    # Define paths
    image_path = 'images/obama.jpg'
    cascade_path = 'cascades/haarcascade_frontalface_default.xml'
    
    # Load the image and convert it to greyscale
    image = cv2.imread(image_path)
    image = imutils.resize(image, 600, 600)
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    
    # Find faces in the image
    detector = FaceDetector(cascade_path)
    face_boxes = detector.detect(gray, 1.2, 5)
    print("{} face(s) found".format(len(face_boxes)))
    
    # Loop over the faces and draw a rectangle around each
    for (x, y, w, h) in face_boxes:
        cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 2)
    
    # Show the detected faces
    cv2.imshow("Faces", image)
    if(cv2.waitKey(0)):
     cv2.destroyAllWindows()
    
    1. Links to necessary files:
      Haar cascade frontal face
      Obama Family Image


  • @arunksoman thankyou 🥳



  • @Nandu But I have to mention that it is not a deep learning method. It is based on Integral images(Viola-Jones algorithm), which is basically something about ML. From opencv 3.4.3 there is a DNN module. This module help us to load caffemodels, torch models as well as tensorflow models. You can find out caffemodels on the Internet in order to detect faces. Using those we can make face detection quite efficiently. If you have any doubt feel free to ask here.



  • @arunksoman how this code helps me to count faces if deeplearning isn't used.



  • @Nandu Please read the comment given above carefully and search how the viola-jones algorithm works. Sorry for misunderstanding what you say. That is why edited comment.



  • @Nandu Did you complete? excited to see.



  • @salmanfaris in the below terminal count shows.some steps i have followed in a different manner.Thank you for helping me!🙂

    IMG-20200314-WA0032.jpg



  • Happy to know that 😊 , thanks to @arunksoman 🤝 , Happy Making 💥



  • @Nandu You are still confusing me because your title saying you have to count faces, your picture shows, counting of cars🤔. Actually what are you trying to make?



  • @arunksoman yes.In my project ,subject is car.So if i get some idea regarding counting faces then i could modify it.In my project i don't want to copy anyone's code,for getting some idea i have asked you.



  • @Nandu Are you using your own haarcascade? If not try to make your own one car.xml. It will be a nice experience. One more doubt, why are you not detecting other vehicles rather than car?



  • @arunksoman how do we make the cars.xml file???



  • @Mennyt Hai Mennyt. There is a couple of ways:

    1. Using Cascade Trainer GUI- Tutorial
    2. Programmatic way

Log in to reply
 

Recent Posts

  • A

    @sreu13 I edited comment please read that again. Most probably it will not effect your file system. In some scenarios it can create worse effect.

    read more
  • S

    @arunksoman i'll try this method, but while executing swap command, will the rasbian os and the files it contains be effected??

    read more
  • A

    @sreu13
    Ensure that you are installed tensorflow 1.x since screenshot of your code shows something like
    from keras.layers.convolution import covolution2D
    It is not correct in case for tensorflow 2.0 since keras api is part of tensorflow itself.

    Another thing is try to expand your file system. It should be do with your own risk.

    sudo raspi-config Navigate to Advanced options Select advanced options and hit enter(I believes tab key is useful here) Choose Expand File System and hit enter finish. Then your pi may prompt to reboot. If it didn't run. Execute
    sudo reboot Then execute following commands one by one: $ sudo apt-get purge wolfram-engine $ sudo apt-get purge libreoffice* $ sudo apt-get clean $ sudo apt-get autoremove Then increase swap memory by editing following file swapfile:
    $ sudo nano /etc/dphys-swapfile It will open nano editor. Navigate to variable CONF_SWAPSIZE=100

    It indicates your current swap is only 100mb.
    So you just have to increase by commenting down this line and increase to appropriate value something like shown below for 2GB swap:

    #CONF_SWAPSIZE=100 CONF_SWAPSIZE=2048 Save the file and exit nano editor. sudo reboot

    I believes if it did not helped you, you have to think about MOVIDIUS or NVIDIA Jetson nano etc.

    read more
  • S

    @arunksoman tensorflow 2.0.0 version had been installed

    read more
  • @arunksoman thanks,
    I will look more into it. Thanks again

    read more