On the afternoon of March 19, 8 teams from five universities entered the main venue.

There are two teams from Fudan University, Jiaotong University, and Tongda University.

That's the benefit of discipline strengths.

Tokyo University and China Normal University may win, but these three universities will never lose, and there will still be a team to participate in the national university competition in the end.

The key thing is that these master's and doctoral students from first-class universities in China are really not boasting, they are all extremely smart, and they all have their own unique insights in artificial intelligence.

In the face of these elite talents, Richard also cheered up 120,000 points.

The topic he chose was naturally machine recognition, which he was good at, and it was precisely because of Princeton University's disciplinary advantages in this area that he was successfully promoted.

Chen Xu, who was playing auxiliary, was sitting in his position drowsy at this time.

He is really not interested in the research of convolutional neural networks in machine recognition applications, Xiaobai has been able to "read and write" since the day of his birth, and if it were not for various reasons, he would have developed more powerful image recognition software long ago.

Jiang Xiaorou next to him had a worse and worse impression of him.

After reading his project report at the beginning, I felt that this person was a manufacturable material, and I was impressed by him, and even let him give a speech on stage.

But what did he do?

There are also games in the past few days, he just can't work, he can't help anything, and he often finds excuses to miss the game, and he doesn't know why he went.

Thinking that some team members said that he used to eat, drink and have fun with the activity funds, Jiang Xiaorou suspected that he just took the opportunity to go out to play.

At this time, looking at him sitting there and drowsy, he couldn't help but pound him, "Teachers and classmates are watching from the audience, if you want to sleep, please go back and sleep again." ”

"Ah...... I didn't sleep ......" Chen Xu immediately sat up straight, reached out and wiped the saliva from the corner of his mouth, and looked around, Richard's body helped him block the sight of the audience and the judges and teachers, and he couldn't see it at all.

Last night, Shen Yiqiu called him to go to the bar to drink, and he didn't go home until two o'clock in the morning.

After arriving home, the two of them video chatted for more than an hour, and went to bed at nearly 4 o'clock.

Of course, the two of them are not in a relationship.

Shen Yiqiu said to treat him as a buddy.

Chen Xu also doesn't like women who go clubbing and drink.

But it doesn't matter if it's a friend.

The game was still going on, and as he sat there, his eyes began to fight again, and at this moment, there was a "beep" sound in the competition venue, which turned out to be the same university submitting the test set data.

Soon Fuda and ECNU submitted them one after another.

Here, Richard, with the assistance of Jiang Xiaorou, also completed the test and was ready to submit the data.

Chen Xu said, "Wait a minute. ”

Richard turned his head: "What's wrong?"

Chen Xu: "Let me recheck." ”

During this period of time, Professor Tang Tianhua has always taken great care of him, reciprocating his favor, and he can't watch the defeat of the University of Tokyo.

Richard turned his head to look at him, a little tempted to laugh, and said, "You are ...... Seriously?"

It's not that he looks down on Chen Xu, but based on the other party's performance in the past few days, he very much suspects that his several project reports were bought with money, and the real level is open to discussion.

Now, he's come to review his data, who gave him the courage?

Ignoring the suspicious gazes of Richard and the three team members, Chen Xu stood up and said, "Come, let go." ”

Richard stared at him, and seeing that his expression didn't look like a joke, he nodded and got up and stood aside.

Chen Xu didn't care what Richard was thinking, and after sitting down in front of the computer, he stretched his fingers and clicked on Richard's test set algorithm system to see possible errors in it.

Machine recognition is the selection of 20,000 images from 2 million images on the Internet, which belong to 40 categories of daily scenes, such as terminals, football fields, etc.

Each scene category contains 300-500 images.

Then, an algorithm is established according to the image scene dataset to predict the scene category to which each image belongs, and the organizing committee will score the prediction accuracy by calculating the error between the predicted value submitted by Tokyo University and the true value of the scene.

Below the average is considered a knockout.

The huge data stream was like a waterfall, and Chen Xu watched it quietly.

Ten seconds later, there was a prompt sound in his ears, and Chen Xu pressed the arbitrary stop button to recompile the test data algorithm system.

Before Richard, who was standing next to him, could stop it, his eyes turned into a brass bell when he saw the matching rate value in the upper left corner of the computer.

The match rate is an early assessment of the predicted value and the true value, and a higher value usually means a higher accuracy rate. Similarly, the matching rate value is also applied to the projects of other universities.

Before, his match rate value only reached 81.36%, but just now after Chen Xu modified his model database, the match rate suddenly became 82.17%, which means that the other party's leak detection is real and effective.

But how is this possible?

Children all know 1+1≠3, but how to demonstrate the steps of 1+1≠3 is the key, and this is exactly what Chen Xu is doing now.

What Chen Xu is doing now is not modifying his algorithm system, but rewriting it.

What's more, the other party didn't even hang up the test software at this time, just with a pair of naked eyes looking for errors and omissions in the huge data stream for modification, this is no longer something that genius can describe, this is completely a pervert!

Richard changed color instantly, and there were huge waves in his mind, and he watched Chen Xu operate there in a daze, and he didn't dare to gasp.

Not only did Richard discover the problem, but Jiang Xiaorou and the other two team members who were about to stop him next to him also discovered that the situation was wrong, and they all gasped.

Including Jiang Xiaorou, a student at Princeton University, in front of Chen Xu, a monster, he has also become a babbling child at this time, which is not worth mentioning at all.

The match rate is still rising: 83.26%;

84.46%；

85.78%；

86.19%……

The students in the auditorium also noticed the situation.

"Hey, look, Dongda seems to have changed generals on the verge of battle. ”

"The game is almost over, and now it's still a fart!"

“……”

The data on the computer is connected to the big screen, and someone soon sees the data changes on it.

Seeing that the match rate skyrocketed, someone shouted, "Look at the big screen! Damn, there's a bug." ”

"Hey, it's really good~ I saw that the matching rate of Dongda was only 81% before, why did it become 86% in a blink of an eye...... Groove ...... 87% up. ”

The highest match rate value in the field before was one of the teams of Fuji, 86.79%, and the other team was 86.5%;

Jiaotong University: 85.55%/84.71%;

Same as large: 83.14%/82.66%;

ECNU: 80.71%.

Professor Tang Tianhua also noticed this scene, first he was flustered, and when he saw Chen Xu sitting there on the stage, he was both surprised and happy for a while, and the young man who could be vigorously promoted by an international academic giant like Liu Chenglin really had a few brushes!

The match rate is still improving.

When the matching rate reached 90%, the jurors of the on-site organizing committee were alarmed, and they all looked at the area of Dongda.

The match rate is 91%;

The match rate is 92%;

The match rate is 93%......

“……”

"Wow............"

The chatter in the hall grew louder and louder, and many people stood up.

Ji Qinglin, Yang Shihan, and Zhou Kuo in the audience of Dongda were all excited, and couldn't help shouting: "The boss is mighty......"

……

……

With the attitude of almost crushing Fuda and Jiaotong University, Dongda won the highest score in the Zhonghai Station area of the Global University AI Challenge, and was shocked.

After the competition, the judges of the expert group, experts and scholars from major universities in China and Overseas inquired about Chen Xu's origin, and after knowing that he was Liu Chenglin's "student", they suddenly realized.

Chen Xu's name quickly spread in domestic academic circles.

At the same time, a sponsor of the competition approached Tokyo University that night with only one purpose - to buy their image recognition technology!

These sponsors are all well-known network technology companies in China, and they see the international leadership of Dongda's image recognition technology and want to buy it first.

Once you can buy this technology, you will definitely make a lot of money.

High-precision image recognition technology has a wide range of commercial application markets.

For example, when laser is doing positioning cutting, positioning marking, and positioning welding, it is very necessary to help high-precision image recognition technology, otherwise, it is difficult to achieve accurate positioning.

There is also a manipulator on the production line when grabbing the workpiece, when it is inconvenient to do hard mechanical positioning, it must be positioned with the help of image recognition technology.

In addition, many automated production lines need the help of image recognition technology.

It can be said that where there is automatic control, when the mechanical positioning cannot be completed, image recognition will show its skills.

This is still only industrial, in addition to navigation, map and terrain registration, natural resources analysis, weather forecasting, environmental monitoring, physiological disease research, e-commerce, enterprise process management and many other fields have very important application value.

The technology used by the participating teams to participate in the competition can be traded as long as it can be applied to business and there are companies willing to buy or invest. And this is also the reason why universities are strongly sponsored.

What is the hundreds of thousands of bonuses, such as meteorology, medical treatment, manufacturing, travel, unmanned driving and other research topics, once they can be commercialized, their value is incalculable.

However, it is a pity that the image recognition technology is not in the hands of Dongda, and they do not know anything about the architecture of the algorithm system.

Of course, this is only the personal statement of Professor Tang Tianhua, the person in charge, and has not received an official reply from the hospital and the school. The school doesn't know about it yet.

The major network technology companies immediately went to the project leader Richard, and the price was 50 million yuan to buy his image recognition technology patent.

When Richard heard this, the adrenaline rushed wildly.

Even if his family is relatively wealthy, 50 million is also a huge amount of money for him.

Richard couldn't care so much, so he immediately returned to the laboratory, downloaded the test set algorithm system used to participate in the competition, and analyzed the algorithm principle.

As long as the core principle can be analyzed, he can say that this image recognition algorithm system is his patent, and the patent has been registered.

Many people know that 75% of the code in this algorithm system was written by him, including the algorithm architecture, which he also put forward ideas, and Jiang Xiaorou helped him improve them.

So even if he preemptively registered, that Chen Xu didn't have any way to take him, and he was not afraid of him when he went to the court to litigate.

As for what is Dongda? It's a big deal to go back to the United States.

With excitement, Richard used the analysis software to tear down and reset the algorithm system.

The progress bar was parsing all the way to 11% without any problem, and everything went well.

But in the next second, the parsing software stuck there and didn't move, a big red exclamation mark popped up on it, and an alarm sounded.

"Toot-toot-toot-toot-"

Richard took a closer look, and the core source code was encrypted......

()