Machine Learning Based Facial Recognition
Humans have an innate ability to recognize faces in nearly any situation, allowing us to differentiate one person from another, regardless of the environment. Local binary patterns (LBP) utilize depth perception to quantify the pixels of a subject’s face depending on the person’s unique facial structure. Through machine learning methodology, multiple faces can be stored in a system over time (known as training the system). As a system encounters more faces, its database expands and correspondingly grows artificially “smarter”, being able to identify more individuals. Using Python, a facial classifier was coded utilizing LBP to collect and store subjects’ faces with their corresponding names. Afterward, a live feed recognizer was coded, which also utilized LBP to convert live frames into grayscale images that could be compared and paired with those stored from the facial classifier. It was expected that the classifier would be about 80% accurate due to the camera quality used during the development phase. Ultimately, exploring recognition systems driven by machine learning can help further research in facial feature detection for analytical purposes.