███████╗ █████╗  ██████╗███████╗        ██████╗ ███████╗████████╗███████╗ ██████╗████████╗
██╔════╝██╔══██╗██╔════╝██╔════╝        ██╔══██╗██╔════╝╚══██╔══╝██╔════╝██╔════╝╚══██╔══╝
█████╗  ███████║██║     █████╗          ██║  ██║█████╗     ██║   █████╗  ██║        ██║   
██╔══╝  ██╔══██║██║     ██╔══╝          ██║  ██║██╔══╝     ██║   ██╔══╝  ██║        ██║   
██║     ██║  ██║╚██████╗███████╗███████╗██████╔╝███████╗   ██║   ███████╗╚██████╗   ██║   
╚═╝     ╚═╝  ╚═╝ ╚═════╝╚══════╝╚══════╝╚═════╝ ╚══════╝   ╚═╝   ╚══════╝ ╚═════╝   ╚═╝   
                                                                                          

Facial Recogination

Face Recogination From the Scratch From Creating Image DataSet to Training and Evaluating the Model.

Requirements

• Google Image Downloader Library- pip3 install google_images_download
• Chrome Web Driver-ChromeDriver - WebDriver for Chrome
• OpenCV- pip3 install opencv-python
• TensorFlow- pip3 install tensorflow
• Keras- pip3 install keras
• Numpy- pip3 install numpy
• Pillow- pip3 install pillow
• imutils- pip3 install imutils
• Shutil- pip3 install pytest-shutil
• Pathlib- pip3 install pathlib
• imghdr- pip3 install micropython-imghdr

Step 1 : Creating Image DataSet using Google Image Downloader

python3 imgdownloader.py

• For example : I want to build a face recogination for All Game of Characters Like:
  • Nikolaj Coster-Waldau =Jaime Lannister
  • Lena Headey =Cersei Lannister
  • Emilia Clarke=Daenerys Targaryen
  • Iain Glen=Jorah Mormont
  • Kit Harington=Jon Snow
  • Sophie Turner= Sansa Stark
  • Maisie Williams=Arya Stark
  • Alfie Allen =Theon Greyjoy
  • Isaac Hempstead Wright= Bran Stark
  • Jack Gleeson=Joffrey Baratheon
  • Rory McCann=The Hound
  • Peter Dinklage=Tyrion Lannister
  • Jason Momoa= Khal Drogo
  • Aidan Gillen=Littlefinger
  • John Bradley=Samwell Tarly
  • Sean Bean =Eddard Ned Stark
  • Michelle Fairley=Catelyn Stark
• We need to Download lots of images of these character and to do that we will run "imgdownloader file"
  • It will ask for user inputs like
    • Enter any keyword or keywords(sperated by comma) for eg. Nikolaj Coster-Waldau,Lena Headey,Emilia Clarke
    • Number of Image for each keywords for eg. 2000 for 2000 images for each
    • Output Folder name Assets/Sample/GOT. full path of the folder where you want to create an output folder
  • More the number of image more time it will take. Note: Number of image and number of downloaded image may vary due to lack of image or due to some error.
  • once Downloads Completed You will find the output folder inside the location you have provided and subfolders with each keyword name inside the output folder.
    • for eg. Assets/Example/GOT/Nikolaj Coster-Waldau
    • for eg. Assets/Example/GOT/Lena Headey
    • for eg. Assets/Example/GOT/Emilia Clarke

Step 2: Data cleansing from the image dataset

python3 faceExtrator.py

• Downloaded Image may or may not contain Relevant face, So we need to Clean it to make our Model More accurate and effecient. To do so.
  • It requires an input which is nothing but the path of the Directory which consist the Dataset for eg. Assets/Example/GOT
  • After that it Will Transverse through each Image file in the Dataset folder i.e Assets/Example/GOT
  • Then it will Convert the image in Black & White , it will Detect a Face and Get the Region Of Interest (Roi) and Save it as new image in the Output folder after Resizing it to 64X64 ( dimension can be change but it should be small )
• After Cleansing we will get the more Accurate data to extract features from it.

Step 3 : Training and Evaluation

python3 modelt&p.py

• First the code will Extract the Labels and Feature from the Image and Save it in a Folder inside the Example Folder.
• Then it will Save Labels and Features in a File.
• Then it will Shuffles the data and Decide the Percentage of Training and Testing Dataset.
• After that it will Start its Training and generate a stuitable classifier for the face recogination.
• It will Save the Classifier in the Output Folder.

Final Step: Evaluation and Detection

python3 face_detect.py

• It has Variety of options Like
  • Detect Through Webcam
  • Detect Through IP Camera
  • Detect Using Video File
• It supports Single Face detection and multiple face detection as well
• Once Face is Detected it will Preddict Name the Person