I saw an interesting question: if a person's DNA sequence was taken as input and a frontal photo was used as output, and thrown into a deep neural network to learn, would it be feasible?
Of course it works, and someone is already doing it. For example, everyone has heard of "genetic recognition people", let's talk to you today.
01, Is the genetic recognition person coming?
J. Craig Venter, the biology god I introduced earlier, had done just that.
The godfather of gene sequencing published the results of face prediction software. He sequenced the whole genome of people of different ages and ethnic backgrounds, and then fed the genetic data into a computer.
The age composition of the study group can be said to cover most age groups.
The ethnic composition of the research object basically covers the mainstream ethnic composition. Like what:
非裔Ancestry components are African (AFR)；
本土美国人Native American (AMR)；
中亚裔 Central South Asian (CSA)；
东亚裔 East Asian (EAS)；
欧洲人European (EUR)
基因测序数据组成Inferred genomic ancestry proportions for each study participant.
Genetic recognition of people, this great innovation, is about to start...
02, how does he operate?
Next, enter the individual's appearance information, such as three-dimensional facial structure, voice features, biological age, height, weight, BMI index, eye color, skin color, baldness or hair color, and other characteristics [1]
The next step is to train it with a newly developed algorithm, which is known as machine learning.
What was the result, did he succeed?
03, what is the result?
Based on their genetic data, as well as high-quality 3D photos of the participants, their algorithm can make very successful predictions about how people will look. The figure below is a plot of pixel changes for gene prediction and the three-dimensional structure of the face
Model training
Let's get straight to the results:
The figure below shows the results of genetic prediction, with the real face of the individual on the left and the genetic prediction on the right
For example, they randomly selected a group of people (10 people) from a public database, Human Longevity (HLI), and the accuracy rate was 74%, which is more than the proportion of most criminal investigations.
You know, there was a study that was done to ask witnesses to describe the appearance of criminals, and the results were dismal in accuracy. For example, I found a random study [2]
ObjectiveTo verify the technology and accuracy of the combination of simulated images and computer portraits of criminal suspects through eyewitness descriptions. MethodsThe hairstyle, chin, eyes and other facial features of 15 male and female photos were removed separately, and one of them was identified as the target (object), which was identified by 786 witnesses. ResultsThe overall accuracy of witnesses in identifying facial features was 17.4%.
The accuracy rate is only 17.4%, which is basically meaningless. Visible genetically predicted cattle forks. It's not just the face, but there are other factors like age
For example: eye color
For example: height, weight, BMI
For example: skin color
He also gave the prediction accuracy of this model, although the accuracy of different indicators varies, but the overall is quite high
This is a real blood recognition.
You know, everyone must be very familiar with it, the ancients were very popular to recognize relatives by dropping blood, of course, modern science has denied this, because there is a great failure rate of blood recognition, and it can also be cheated through technology, so poor ancient people who are not recognized...
04, What is the use in real life?
You must be familiar with finding criminals through the DNA of criminals' blood traces, but it's actually quite difficult, because you have to find a match, otherwise, even if you get the criminal's DNA, you can only wait stupidly for the day in the future, you are lucky enough to find the criminal's DNA again, and then the two are compared online. If you can't collect the criminal's DNA, you'll never find the criminal through genetic evidence.
However, this use of DNA to directly predict a person's appearance is much taller, and it also directly leads to a core problem: privacy.
So at that time, this article caused a lot of controversy as soon as it was published. Some felt that the article was wrong, and some felt that it violated privacy. But in any case, it was a huge discovery.
In the future, with more and more samples and more and more adequate training, it is very possible to predict people directly through DNA.
So:
Are you still hesitating whether the other party has had plastic surgery?
Still worried about becoming a different person after removing your makeup?
Are you still speculating about whether the other party used the legendary PS artifact?
For a long time, the three major evil arts in Asia, makeup and plastic surgery, have plagued countless people, making everyone very confused about a person's true appearance. Maybe one day, this won't be a problem anymore!
【1】Lippert, Christoph, et al. "Identification of individuals by trait prediction using whole-genome sequencing data." Proceedings of the National Academy of Sciences114.38 (2017): 10166-10171.
[2] Research on the ability of witnesses to distinguish physiognomy features[J]. Forensic Science And Technology, 2001(05):20-27.