Top of the Great Era

Chapter 1334 Human Eyes and Mosaics
Superman is a pure technical novice, the kind who doesn't understand anything.

Although Zhou Buqi and Pei Yao are also technical laymen, as corporate executives, they still need to know certain technical theories, and they are far better than ordinary people.

I am very excited about the concept of doubling the image size.

That's great!

It is no wonder that Professor Tang Xiaoge vowed that in the past few years, the scientific research level of his multimedia laboratory is comparable to that of MIT and Cambridge.

This "picture enlargement" project is the hard core of technology!

Zhou Buqi even had a bold guess at a certain moment.

This multimedia lab...

Could it be the famous artificial intelligence laboratory SenseTime?
Artificial intelligence is a high-end and cutting-edge scientific field. Even if the Hong Kong school is very good, it cannot be played casually.Moreover, the scale of papers has reached the level of MIT and Cambridge, which is not something that can be achieved overnight.

The moment he saw the enlarged picture, Zhou Buqi made up his mind.

This laboratory must be funded, supported, and invested!
99.99% of this laboratory is the future SenseTime!

The so-called picture enlargement is actually very simple, just like Superman's cognition, simply stretch the side.

But when zooming in, the problem arises.

As the image is enlarged, the resolution decreases.

A 720P picture becomes 360P when it is doubled, and then becomes 180P when it is enlarged again, and then it becomes a mosaic piece by piece.

Professor Tang's project is to maintain the original resolution while enlarging the picture.

It makes so much sense.

A doctor named Li Jinzhong from Professor Tang came to the stage to explain the project to the projected pictures on the big screen——

"A lot of old photos in the past will be blurry after a long time, or the shooting equipment is not good. Through this algorithm program, the photos can be cleared."

"If it is further developed, it can even restore the color. Turn the old black and white photos into color photos. Even the blurred text in a photo can be recognized more clearly."

"Mr. Pei just mentioned Meitu Xiuxiu. The teacher is not very clear, but I know. Our algorithm can provide a great help to Meitu Xiuxiu. Users can take a photo first, and then choose to use it Which point to focus on. You can also use filters to turn the picture into a variety of styles.”

Zhou Buqi is very satisfied.

Pei Yao was also very satisfied, smiling all over her face, and asked briskly: "Dr. Li, are the ideas you mentioned, color restoration, and different styles of picture filters, can these be realized now?"

Dr. Li Jinzhong was a little embarrassed, "Not yet, we have not conducted research in this area. However, the technical principles used are generally the same as the way artificial intelligence is used, similar to image enlargement, and they all belong to image processing of deep learning. "

Professor Tang next to him was more introverted, and said seriously: "To realize these points, we need a deeper level of artificial intelligence learning. It is not enough at present, and we need the joint efforts of colleagues from all over the world."

Zhou Buqi said with a smile: "Professor Tang is free, I hope you can take the team to the capital to communicate with our CTO, Dr. Shen Xiangyang. Our Ziweixing Research Institute is focusing on artificial intelligence, and even established a In the AI ​​laboratory, one of the most important directions is the optimization and design of the deep learning algorithm framework."

Professor Tang was very happy, "If there is a chance, it would be the best."

Little Superman just had a brief communication with Ziweixing's experts in private, and now he understands that this algorithm is to improve the resolution of the picture, so he asked curiously: "I remember that many Hollywood movies act like this, what? "Mission: Impossible", "007", "Bourne Shadow" and the like, it is said that the faces of people on the street were captured through satellites, and the pictures were very blurry. Then they were processed through technology to supplement the details and become a picture Complete face, and then dock with some system, you can find all the information of this person."

Professor Tang laughed, "Yes, a lot."

"is that true?"

"fake."

Little Superman was a little disappointed, "Oh."

"Can you be specific?"

But Zhou Buqi was reminded by what Little Superman said, thinking it was very important.

Ziweixing is working on a project called Micropoint Map, and the target is Google Maps.

Google Maps uses commercial satellites from the United States to take pictures of the earth and transmit them to computers for data processing and integration. It becomes a Google Map that is so precise that it leaks secrets. You can even see the flowers and plants in the community on it.This is mainly due to the strength of technology. The precision of commercial satellites in the United States is higher than that of domestic military satellites.

This has endangered national defense security and will be photographed in many confidential places.Americans don't know what these secret places are, so it's useless to photograph them.

However, after "Google Earth" and "Google Maps" were launched, the "Mark" function was provided.

In other words, whoever knows what these places are can mark them.Many people in the country don't know why, so they put it on the mark.Or when some bad guys collect money, they sneak over to investigate, and mark it after the investigation is clear, which is suspected of leaking secrets.

Therefore, the domestic approach is to restrict the use of related products on the one hand, and to launch similar products that are domestically controllable on the other hand.This important task related to national security is mainly undertaken by Baidu Maps and Micropoint Maps.

However, what should I do if the precision of domestic commercial satellite photography is too poor?
The National Geographic Bureau and other relevant departments have organized some civilian or military helicopters to take aerial photos of big cities.The precision of this kind of aerial photography is much higher than that of satellites.

But the problem came again.

The country is so big, how can it be possible to use aerial photography?How can there be so much money, how can there be so many manpower, material resources and aircraft? (In fact, it has not been solved yet, and it will be done by blocking Google once and for all)

It was Zhou Buqi who solved this problem. He took the lead in contacting the National Geographic Bureau and Bajixing, ordered a batch of drones, and used them to take pictures.

It is a very meaningful strategic non-profit project to record the ever-changing development and changes of the country.

But here's another problem.

The technology of drones is not mature enough, and it may not be stable enough when shooting, which will lead to out-of-focus and blurred pictures, and even many details problems due to sunlight, exposure and other reasons.

Aerial photography by drone is definitely not as good as people sitting on a helicopter with a professional camera with a telephoto lens.

The current research project in the multimedia laboratory seems to be able to solve this problem!

The Multimedia Lab, as the name suggests, researches "projection", "imaging", and "picture processing" laboratories. In this field, they are more professional than artificial intelligence expert Shen Xiangyang.

Through the artificial intelligence algorithm, the photos taken by the drone are processed and turned into clear photos, which can be similar to the effect of artificial aerial photography.

Because this is a Ziweixing project, Micropoint Maps has a technological monopoly, and Micropoint Maps can suppress Baidu Maps.

It's so much!

What's more, it can even be extended to the military field.

For example, in military investigations, the blurred pictures captured are cleared; for example, foreign pictures captured by domestic satellites are too blurred, and they can also be cleared.

However, there is obviously still a long way to go. Professor Tang said: "During the investigation process of the police and FBI, some blurred photos of criminal suspects are often obtained. If the photos can be super-resolution processed, the If the photos become clear and beautiful, the rate of solving crimes will be greatly improved. But in fact, the United States has been doing this for decades, and it has not been able to do so far.”

Zhou Buqi asked, "Why?"

"The problem with the algorithm," Professor Tang sighed, "We are very disappointed with the algorithm in the past."

Zhou Buqi asked: "Can deep learning based on artificial intelligence solve it?"

Professor Tang said: "This matter is very complicated. The computer processes two signals of 0 and 1. It will be very complicated to complete some human behavior judgments through these two signals. For example, there is a table here, which we can see at a glance. It can be judged without reaction time. However, if a computer wants to recognize a table from a picture, it may need millions of calculations. In fact, in the field of picture recognition in the past few decades, the processing efficiency of computers is far Below the human eye. What the human eye cannot see clearly, the computer is even less likely to see clearly."

Zhou Buqi seemed to understand, "In the past, the processing power of computers was too poor, but now there are many supercomputers, Ziwei Star can even perform clustered cloud computing, and the computing power has been greatly increased."

Professor Tang's eyes on Boss Zhou changed a little, and he said happily: "Yes, the improvement of computing power is the basis for the development of artificial intelligence. However, the previous algorithm is definitely not good, and there must be new ones. Innovation and new discoveries in academics. Let me give you an example. If 1 is magnified by 2 times, it is 2; if it is magnified 100 times, it is 100; All are exact numbers. However, for an uncertain number ε, if it is magnified 1000 times, it is 1000ε, and if it is magnified 1 times, it is 10ε. No matter how many times it is magnified, the result is an uncertain number.”

Those present are all highly educated, and even laymen can understand the logic behind it.

Just like this portrait picture demonstrated on the big screen.

The original photo of this picture is clear, which is a definite number, so no matter how many times it is enlarged, it is based on an algorithm and is definite. After zooming in, a clear image can be given.

But if the original photo is too blurry, so blurry that even the eyes can't see it clearly, it is an uncertain number.No matter how many times it is magnified, it is uncertain.

Zhou Buqi nodded, feeling more at ease about it.

Even if the photos taken by drones are blurred, they can still be seen clearly by human eyes.If the human eye can see clearly, it should be able to handle it well.

Professor Tang went on to say: "However, under the new technology, uncertain numbers that cannot be distinguished by the human eye may not necessarily be transformed into definite numbers!"

"Oh?"

Zhou Buqi's eyes lit up.

Professor Tang said: "Rely on big data."

(End of this chapter)