Face biometrics #

The class covers face detection, face recognition and methods of spoofing them with deep fakes.

Pre-read (Required) #

• https://www.tensorflow.org/guide/keras/sequential_model
• https://scikit-image.org/docs/dev/auto_examples/features_detection/plot_hog.html
• https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_objdetect/py_face_detection/py_face_detection.html

Class content #

• Introduction to face detection algorithms (Histogram of Oriented Gradients, Haar Cascades, Deep Neural Network). The aim of the task is to get to know the methods and their strengths and weaknesses.
• Implementation of the HOG (Histogram of Oriented Gradients) algorithm to detect characteristic points of the face.
• Development and implementation of selected features and algorithms to verify the similarity (e.g. calculating the distance between face elements and their comparison).
• Implementation of the identification function based on the face pattern. A test set is provided for the task, on which the effectiveness of the solution should be verified. The function and algorithms used in this activity will be evaluated in the next class. The group with the highest score will receive additional points.
• Preparation of a set of training data for eye color recognition and model implementation.
• Comparison of the deep learning method with the traditional, previously implemented methods.
• Analysis of the safety and effectiveness of the implemented methods.
• Deep fake implementation and its practical applications.