- Trending Categories
- Data Structure
- Operating System
- C Programming
- Selected Reading
- UPSC IAS Exams Notes
- Developer's Best Practices
- Questions and Answers
- Effective Resume Writing
- HR Interview Questions
- Computer Glossary
- Who is Who
Face Detection and Recognition Systems
10 Lectures 1 hours
15 Lectures 54 mins
Imagine standing in front of the entrance door to your office and see your face flash up on a large screen. If the screen identifies you, it opens the doors; if it doesn’t the doors don’t open. Your face is your digital ID. It allows you to enter your office and authenticate your identity in a hoard of other places. Sounds uncanny, but is undeniably remarkable! This not so old technology is on its way to becoming ubiquitous.
Enter the world of face detection and recognition systems…recognizing the face from all angles, with even more accuracy than a person can. Using machine learning and other deep learning techniques, computers can recognize faces with utmost accuracy. The facial recognition systems comprise of high-end hardware components along with proficient software for identification and verification of a person by comparing the facial features from the person present to the features stored in the facial database. The facial detection and recognition system can verify a person from a digital image or a frame/clips of a video. Numeric codes called faceprints are used for detection along with identifying 80 nodal points (end to end measurements with respect to nose, jaws etc.) on the face.
Web and desktop applications can also benefit from such facial recognition systems to avoid hacking. Enforcement agencies can use this for locating an individual in a crowd. Companies can keep track of employees attendance. These systems are already being used by agencies targeting criminals, at airports for security or to differentiate players during games. Marketing personalization makes use of billboards with software which identify the demographics of passersby for targeted advertising.
No retraining − The training is usually done on the face and not to the existing images in the database.
Quick induction − As the new face is added to the database, it should not take more than 0.01 sec for the addition along with the training, made possible with the use of a single core processor.
Self-supporting Feature Database − Post the training, the facial feature database must be different from the face image database. This is usually done to protect the (possibility of) tampering of images in the facial database. Distance between jaw lines, nose tips, lips contours, eye centers are all matched during face detection and recognition process.
Resolution dependent − The images or the face recognition algorithm must be independent of high resolution images; the videos however must be at or higher than 640*480 resolution.
Robust − The face recognition system must be sturdy and not weak towards variations in posture, facial expressions and luminescence (albeit within a certain limit).
Face changes − Beards, moustaches and other such features can make a difference to the actual image. Hence, the facial database must include the original image as well as one with the additional features with the identical ID. Eyes and eyebrows are usually also used as signatures of the image.
Other aspects − Eye glasses must be clear so as to reveal eyes clearly.
Scalability − The images in the facial database must be scalable and must have the capacity to store more than 1000 images. The format could be in many versions such as JPEG, BMP, PNG etc.; whereas the videos could be in MJPEG, MPEG4 etc.
Accuracy − The face recognition must work at an extremely high accuracy rate (more than 90%).
Forensics − The image and the video must be watermarked as well as encrypted for forensics, making it easier for authentication purposes.
Finally, the database must be integrated and include demographics such as name, age, sex, date of birth, nationality, address, employment history, unique identification number etc. And the query for any person must not take more than five seconds in a database of 5000.
Difference from Biometrics
Biometrics uses the touch of a person (finger print/palm print) and other features such as iris, but facial recognition is a non-contact and non-intrusive process. Here, interaction with a person is not required; however it goes on to identifying a person with utmost accuracy. And just in case, facial recognition systems don’t work, skin bio-metrics is being worked on as the next step in technological advancement.
Where is it being used most?
It’s China. One can log in, or simply enter a store and pay for things using the face recognition system, authorize payments and other financial transactions, provide entrée to services, surveillance and security systems and most importantly track down criminals.
The key players in China are Face++, Baidu and Alibaba. The system can even track the individuals’ movements all along in a room. Face++ is a startup but the technology is already being used by several organizations. Through Alipay one can transfer money; via Didi, a ride hailing company, one can know if the driver behind the wheel is a genuine one. And this is further ensured with the application requiring the person to move their heads during the scan (to avoid duping with a photo).
Baidu, China’s most popular search engine, is making it easier for people to pick their rail tickets while using the face recognition system. Also, Baidu is pairing up with the government of Wuzhen which is famous for many historic monuments, to allow people to enjoy the attractions without a ticket. This is being made possible through Baidu scanning millions of faces through their database. Furthermore, many housing complexes are using the facial recognition systems to provide access; and shops and restaurants are working towards making the customer experience effortless.
While China has a supremely large database of ID card photos, half of United States population is a part of its facial recognition database. Microsoft is using the facial recognition systems for authentication in Windows 10; Apple and Google use it for tagging and sharing photos. Apart from some smart phone applications using this technology, Facebook also uses it for tagging photographs. However, as per a research, people and in UK and US find the facial recognition system for payments in stores quite creepy and would not prefer it.
The human face is an identity to a person. As the facial recognition system is used for law and non-law enforcement, people might have to give up a bit on their privacy, security and transparency. Hence, till such issues are resolved, the debate over the validity of its usage is going to persist.
- Difference between Facial Recognition and Face ID
- In Google Photo Gallery, Enable Face Recognition
- How Apps like Zcash, Blockstack, and Appcoins are changing the Face of Digital Payment Systems?
- What are the types of Intrusion-Detection Systems in information security?
- Signals and Systems: Classification of Systems
- Signals and Systems: Linear and Non-Linear Systems
- Signals and Systems: Invertible and Non-Invertible Systems
- Signals and Systems: Linear Time-Invariant Systems
- Signals and Systems: Time Variant and Time-Invariant Systems
- Signals and Systems – Filter Characteristics of Linear Systems
- CONVERSATIONAL INTERFACES – Speech Recognition Technology
- What is the Revenue Recognition Principle?
- Audio processing using Pydub and Google Speech Recognition API in Python
- Error Detection and Correction in Data link Layer
- Device Detection and Responsive Design in React JS