Have you ever wondered how the face scanners are managing to detect and recognize your faces even when you keep changing your hairstyle, beard, and make-up? What exactly is face recognition and what is machine learning at all?
Machine learning is one of the major applications of artificial intelligence (AI). It is an ability of the system to learn on its own by its previous experiences without explicitly being programmed. It develops a computer program that can access data and use it to learn different aspects of the data for themselves. The roots of machine learning are in statistics, which can also be thought of as the art of extracting knowledge from data.
The implementation of face recognition has seen many iterations starting from its origin in the 1960s when it was manually implemented with a RAND tablet. However, the adoption of facial recognition on a large scale became possible, thanks to the breakthrough of deep learning in the early 2010s. The facial recognition market was at 3.97 billion USD in 2018 and is predicted to grow to 10.15 billion USD by 2025.
Facial recognition is an application that analyzes and verifies a person by comparing unique facial contours. A face recognition system uses facial landmarks like the eye, nose, lips, etc. to map facial features using an image or a picture from a video frame. It compares the information from a database to find the exact match. Face recognition helps verify someone’s identity and is a widely used security mechanism. Face Recognition is used in many areas like Mobile device owner authentication, school security, airline customs, entertaining apps, etc.
The functioning of Facial Recognition is as follows:
- Face Detection
The camera can accurately detect if the person is looking straight to the camera, even though modern technology allows detection when the person is on their own. To begin, the camera will detect and recognize a face, either alone or in a crowd.
- Face Analysis
After detecting a person’s face, the system captures the face for analysis. The face will be separated into distinguishable landmarks, also called nodal points. Distinguishable landmarks or nodal points make up each face. Facial recognition software will analyze the nodal points such as the distance between your eyes or the shape of your cheekbones. Mostly, the face recognition technology uses 2D images instead of 3D because 2Ds are more correlated with public photos in the database.
- Conversion of the image into data
After analysis, the nodal points of the face become a number in the application database. The analysis of the face is turned into a mathematical formula and the numerical code is called a faceprint. Similar to the unique structure of a thumbprint, each person will have their own unique faceprint.
- Finding a Match
Finding a match is the final step of the process. The facial code is obtained after image to data conversion is compared against a database with photos of their faceprints. The number of faces that are compared depends on the database and how many databases the software has access to. An example of a database that many have access to is Facebook’s photos. Any photos that are tagged with a person’s name become part of the Facebook database. The facial recognition technology identifies a match for the exact facial features – it returns the user with the found match and other relevant information.
How does Facebook’s face recognition algorithm work?
Facebook began using facial recognition in the US in 2010 when it automatically tagged people in photos using its tag suggestions tool. The tool scans a user’s face and offers suggestions about who that person is. And in 2019, Facebook made the feature opt-in as part of its new drive to become more privacy-focused. Facebook uses the Deep face Algorithm. It is trained on a large dataset of faces, an identity labeled dataset of four million facial images belonging to more than 4,000 identities, acquired from a population vastly different from the one used to construct the evaluation benchmarks. In other words, each identity had an average of a thousand samples. The DeepFace algorithm first aligns a face so that the person in the picture faces forward, using a 3-D model of an “average” forward-looking face. Then it uses deep learning to find a numerical description of the forward-looking face. If two different images have similar enough descriptions, Deep face decides they must show the same face.
Concerns Of Facial Recognition
-Error in Identification
Due to poor images or a lack of information in the database, face recognition might fail in accurately matching the faceprints. Different factors like lighting, image quality, data size, etc can trigger the accuracy of the system.
-Privacy
This technology can track you and make your personal information easily accessible. Data breaches are also too common lately. A recent breach was discovered in a UK biometric system used by banks, in which over a million people’s fingerprints and facial recognition information were discovered on a publicly accessible database. This is causing privacy issues in a large mass.
-Data Misuse
Pew Research Center found that 56% of Americans trust law enforcement to use facial recognition responsibly, others aren’t as confident that the data will be used ethically. This clearly shows they have control over the data and the databases are public so even people with malicious intent can find anyone on the database and track them down.