As the number of services using AI technology increases, many people are probably interested in “AI x face” services such as face diagnosis apps.
Also, since face recognition is used for security, have you ever been interested in AI face recognition technology?
Therefore, this time, we will introduce an AI service related to “face”.
Table of Contents
- What is AI
- AI x facial technology
- AI by machine learning x facial technology
- Deep learning AI x facial technology
- AI by GAN x facial technology
- A case of AI x face that became a hot topic
- face recognition system
- Automatically generate the face of a fictional idol
- Face photo creation site spreads
- Automatically generate anime characters
- 3 face diagnosis apps that use AI
- ① Face tag – AI face diagnosis app
- (2) Look-alike – AI diagnoses celebrities who look alike
- ③ Animal face diagnosis
- Face exchange / face swapping is possible with AI
- Face-swapping apps become a hot topic for teens
- Deepfake exploitation damage spreads
- Summary
What is AI
AI is an abbreviation for Artificial Intelligence, which is translated as artificial intelligence in Japanese.
Machine learning, deep learning, and GAN are essential elements to explain AI x facial technology.
Simply put, machine learning is one elemental technology of AI, and deep learning is one elemental technology of machine learning.
Machine learning is roughly divided into “supervised learning”, “unsupervised learning” and “reinforcement learning”, but GAN corresponds to ` ` unsupervised learning ”.
AI x facial technology
“AI x face” technology refers to a face recognition system that uses AI.
A facial recognition system is a set of algorithms that can find, capture, store, and analyze facial features to match images of individuals in existing databases.
An AI-powered facial recognition system can instantly search a database of faces and compare one or more detected faces.
Face recognition systems equipped with AI have the following features.
- Real-time identification possible
- Spoofing prevention measures
- Less bias due to race and gender due to model training for millions of faces
- Information can be shared between multiple cameras
AI by machine learning x facial technology
The face recognition system based on machine learning can be explained in three stages.
- face detection
- face alignment
- face recognition
I will explain each.
Step 1: Face detection
The machine must first locate the face in the image or video.
Today, most cameras have a built-in face detection function, and face detection technology is familiar to us.
Also, social media platforms such as Snapchat and Instagram use face detection to add effects to photos and videos that users take with the app.
Step 2: Face Alignment
When a face is out of focus, the computer sees it as something completely different from what it really is. I need an algorithm to normalize the faces to keep them consistent with the faces in the database.
One way to achieve this is through the use of facial landmarks. For example, under the chin, over the nose, outside the eyes, and various spots around the eyes and mouth.
Step 3: Face Recognition
Using each face measurement, ultimately AI powered by machine learning matches face measurements against known faces in the database.
Determines which face in the database is closest to the dimensions of the given face data and returns that data as the matching face.
Deep learning AI x facial technology
Feature measurement and extraction
Various features must be measured and extracted from the face in order for the algorithm to match it with other faces in the database.
While it was initially unclear what features to measure and extract, researchers discovered that the best approach was to let the machine learning algorithm decide what measurements to collect.
This process is called embedding. It uses deep learning to train itself to generate multiple measurements of a face, allowing it to distinguish faces from other faces.
AI by GAN x facial technology
Manipulating facial images
GAN stands for Generative adversarial network, which translates to Generative Adversarial Network in Japanese.
In recent years, it has emerged as an effective image manipulation tool and is widely used.
It is also possible to manipulate facial images using GAN, and tools such as DeepFake have already been exploited.
A case of AI x face that became a hot topic
From here, we will introduce the “face recognition system” and “automatic face generation of fictional idols” as examples of AI x face that have become a hot topic.
face recognition system
A representative technology for combining AI and face is the face recognition system.
The Tokyo 2020 Olympic and Paralympic Games (hereafter, the Tokyo 2020 Games) have become a hot topic for the first time in the history of the Games that face recognition systems have been introduced.
In the past, people involved were sorted out only by their ID cards, but in the event of lending, borrowing, or theft, there could be cases of unauthorized entry.
It is no exaggeration to say that the reason why there were no major incidents at the Tokyo 2020 Games was the facial recognition system.
Recently, in addition to face recognition systems, face creation and face swapping technologies are progressing and becoming widespread.
Automatically generate the face of a fictional idol
DataGrid Inc. has developed an AI system that generates the faces of non-existing idols.
Recently, there are many cases where contracts are canceled due to scandals.
Using AI-generated talent models has the advantage of reducing risk.
Face photo creation site spreads
Former Uber software engineer Phillip Wang has created a website called “This Person Does Not Exist”.
This website is a site that displays a headshot of a non-existent person every time it is updated.
All photos posted on this site are created using a special artificial intelligence algorithm called GAN (Generative Adversarial Network).
Automatically generate anime characters
Preferred Networks Inc. has developed Crypko, a web service that automatically generates anime characters.
With this service, it is possible to generate an unlimited number of anime characters using AI deep learning.
3 face diagnosis apps that use AI
Next, we will introduce an app that allows you to experience facial diagnosis using AI. The following three applications are introduced this time.
- Face tag – AI face diagnosis app
- Look alike – AI diagnoses celebrities who look alike
- Animal face diagnosis
I will explain each one in detail. Please try it out and see what kind of diagnostic results you get.
① Face tag – AI face diagnosis ap
Quote: App Store – Apple
| price | Free (500 yen for premium version) |
| URLs | App Store-Apple |
▼ Face tag – features of the AI face diagnosis app
- AI analyzes and tags facial impressions
- Tell me about similar celebrities and singers
- Measure the ratio of parts for reference of makeup
(2) Look-alike – AI diagnoses celebrities who look alike
Quote: App Store – Apple
| price | Free (VIP plan is 980 yen per month) |
| URLs | App Store-Apple |
▼ Look-alike – Features of AI diagnosing celebrities who look alike
- Tell me about similar celebrities and singers
- The similarity is expressed as a percentage
③ Animal face diagnosis
Quote: App Store – Apple
| price | free |
| URLs | App Store-Apple |
▼Features of Animal Face Diagnosis
- Recognize a face by photographing it with a camera
- Several similar animals are displayed
- Expressed as a percentage of similarity
Face exchange / face swapping is possible with AI
Face-swapping apps become a hot topic for teens
Reface
Quote: App Store – Apple
| price | Free (with subscription) |
| URLs | App Store-Apple |
▼Features of Reface
- You can replace the captured face with the face of the person in the video or image
- There are a wide variety of replacement materials, including famous scenes from movies.
You can also find applications using AI here.
Face diagnosis sites and apps that utilize AI are introduced here.
Deepfake exploitation damage spreads
Deepfake is a technology that uses deep learning, one of the machine learning algorithms, to combine parts of two images or videos to create a video that is different from the original.
Applications like Reface have also appeared, making it easier to create composite images, but the abuse of deep fake technology has become a social problem.
For example, in April 2021, a photo of Chief Cabinet Secretary Kato’s expression at a press conference after the earthquake was processed to look like he was laughing using a deepfake was circulated.
Regulating processed images is difficult due to the freedom of expression, but it is important to promote legal development.
Summary
In this article, in addition to introducing AI-based face recognition, face creation, face exchange technology, and a face diagnosis app, we will also introduce the essential technical elements of AI x face technology: machine learning , deep learning, and GAN. also explained.
AI x facial technology will improve accuracy and expand services, making it one of the more familiar technologies.
It goes without saying that unauthorized use of photographs that can identify an individual may constitute an infringement of portrait rights.
In addition, publishing photos and videos (so-called selfies) taken by others without permission may constitute copyright infringement.
