Face-Detection

Face detection, also known as facial detection, is a computer technology that uses artificial intelligence (AI) to find and recognize human faces in digital photographs. Face detection technology can be used to enable real-time surveillance and tracking of people in a variety of industries, such as security, biometrics, law enforcement, entertainment, and personal safety.

Steps

1. Collect a video of the persons you want to add into face dection system of atleast - 20 secs

2. Read the video frames

      import glob as gb
      # Reading the video files
      folders = gb.glob('Video')
      list = []
      for folder in folders:
          for i in gb.glob(folder + '/*.mp4'):
              list.append(i)

3. Extraction of image frame Using open-cv extract the image frame from the videos

      # Capture the video
      current_frame = 0
      video = cv2.VideoCapture(frame_list[idx])
      while(True):
          _, frame = video.read()
   
          # Handel the exception error
          try:
              train_image = cv2.resize(frame, (256,256), interpolation=cv2.INTER_AREA)
          except:
              break
   
          cv2.imshow("Train", train_image)
   
          # Saving the image to its corresponding dataset folders
          cv2.imwrite(directory+str(current_frame)+'.jpg', train_image)
          length.append(current_frame)
          current_frame += 1
   
          interrupt = cv2.waitKey(10)
          if interrupt & 0xFF == 27: 
              break
      video.release()
      cv2.destroyAllWindows()

4. Image augmentation Image augmentation artificially creates training images through different ways of processing or combination of multiple processing, such as random rotation, shifts, shear and flips. Below are some image augmentation example:

   Inverting

      def invert_image(image, channel, count):
         image = (channel-image)
         cv2.imwrite(path + "/Invert/" + str(channel) + '-' + str(count)+ ext, image)

   Adding Light

      def add_light(image, gamma, count):
         inverseGamma = 1.0 / gamma
         table = np.array([((i / 255.0) ** inverseGamma) *
                           255 for i in np.arange(0, 256)]).astype("uint8")
         image = cv2.LUT(image, table)
         if gamma >= 1:
             cv2.imwrite(path + "/Light/" +
                         str(format(gamma, '.1f')) +'-'+ str(count) + ext, image)
         else:
             cv2.imwrite(path + "/Dark/" +
                         str(format(gamma, '.1f')) +'-'+ str(count) + ext, image)

   Adding Light Color

      def add_light_color(image, color, gamma, count):
         inverseGamma = 1.0 / gamma
         image = (color - image)
         table = np.array([((i / 255.0) ** inverseGamma) *
                           255 for i in np.arange(0, 256)]).astype("uint8")
         image = cv2.LUT(image, table)
         if gamma >= 1:
             cv2.imwrite(path + "/Light_color/" +
                         str(format(gamma, '.1f')) +'-'+ str(count) + ext, image)
         else :
             cv2.imwrite(path + "/Dark_color/" +
                         str(format(gamma, '.1f')) +'-'+ str(count) + ext, image)

To get started with the project.

   git clone https://github.com/kinley2-glitch/face-detection-using-deep-learning

Authors

• kinley2-glitch