# Facial recognition

> AI that identifies or verifies a person by comparing their face with stored ones.

- Level: Basic
- Topics: AI, Security
- URL: https://tidbit.legios.com.ar/facial-recognition/
- In Spanish: [Reconocimiento facial](https://tidbit.legios.com.ar/es/facial-recognition/)

**Facial recognition** is software that finds a face in a photo or video and works out whose it is. It does not compare pixels. A neural network turns the face into a short list of numbers, a kind of mathematical fingerprint, so that photos of the same person end up with similar numbers despite different lighting, angles or haircuts. Matching is then a question of how close two of those lists are.

There are two very different uses. **Verification** (one to one) answers "is this the person they claim to be?": unlocking your phone, or an airport gate that compares you with your passport photo. **Identification** (one to many) answers "who is this?" by searching a database of many faces, for example a police system checking a security-camera still against a file of photos.

Identification is where the controversy is. Accuracy has improved a lot, but errors still happen: tests by NIST, the US standards agency, found that many algorithms make more mistakes with some demographic groups than with others, and false matches have led to wrongful arrests in the United States. Cameras can also identify people in a crowd without their knowledge. That is why the EU AI Act sharply restricts real-time facial recognition by police in public spaces, several cities have banned or limited its use by government, and European privacy regulators have fined companies that scraped photos from the internet to build face databases.

## An analogy

A doorman with a perfect memory for faces who has studied millions of photos. Checking your face against your own ID card is easy for him; picking one stranger out of a stadium crowd is much harder and much more likely to go wrong.

## Keep in mind

- A face becomes a list of numbers that is compared with others.
- Verification (is it you?) is different from identification (who is it?).
- Error rates can vary across age, sex and skin tone.
- Its use in public spaces is restricted or banned in many places.

## Go deeper

- [Facial recognition system on Wikipedia](https://en.wikipedia.org/wiki/Facial_recognition_system)
- [Face technology evaluations (NIST)](https://www.nist.gov/programs-projects/face-technology-evaluations-frtefate)

## Related terms

- [Neural network](https://tidbit.legios.com.ar/neural-network/): A pile of tiny math units that learns patterns from examples.
- [Database](https://tidbit.legios.com.ar/database/): An organized store of data that programs can search and update safely.
- [Algorithm](https://tidbit.legios.com.ar/algorithm/): A precise, step-by-step method for solving a problem or doing a task.
- [AI regulation](https://tidbit.legios.com.ar/ai-regulation/): The laws that set what AI systems may do, and who answers when they go wrong.
- [Internet](https://tidbit.legios.com.ar/internet/): A worldwide network of networks that lets any connected device reach any other.