Meng Fanqi and Wang Kai have communicated a lot during this time, and they are already quite familiar.

However, seeing this sentence, Meng Fanqi still felt quite nonsensical: "You are a muscular man of more than one meter and eighty, what should I take you away?" Shall I take you or will you carry me on my shoulders? ”

"Brother has been arranged to come to you again." Wang Kai said and sent Meng Fanqi a photo of the airport: "I just arrived at the airport, and the plane at night went to Shanghai. ”

"What's going on?"

"Isn't this that the leader is not at ease, and he doesn't dare to summon you to Yanjing, so he can only send my junior brother to Shanghai." Wang Kai also couldn't help himself, although he had just made a contribution, he always felt that this work had not improved.

"The leader said, let me understand this part of the code by your side, and then go back and explain it to everyone."

"It's still the same work as last time." Meng Fanqi knew that he was not a person of whiteness after all, and Li Yanhong was in a hurry to get government projects, and he definitely needed to send someone to hand over the code to him.

Just looking at the last one, it is estimated that it is not reassuring. Therefore, he sent Wang Kai, who had been communicating with him before, to follow him, and grasp every minute and second to thoroughly grasp this part of Meng Fanqi's new development.

"Okay, it's nothing, we'll see you tomorrow at the whiteness branch near Hechuan Road."

"No problem, brother." Wang Kai agreed very happily: "Brother, if you are useful in the future, you must think of me, no matter what time or place, my brother will definitely do it with you." ”

"What do you mean by taking me away for a long time?" Meng Fanqi came back to his senses.

"That's for sure, there must be no big future for a part-time job, and I can't save much money in my life." Wang Kai didn't hide it, and directly said his truest thoughts.

"I don't remember where I saw it, saying that people can encounter at least one opportunity to change the fate of their lives every day in their lives, but most of the time they don't know it at all."

"You have to change your life to see how you change it, and going out to be created by a mud truck is also changing your life." Meng Fanqi has heard another version: "The version I have heard is that a person will definitely encounter an opportunity to get rich overnight in Silicon Valley for a lifetime, commonly known as the Five Ones." ”

"I think you are my chance, brother, if you really want to start a business in the future, you must call me. The younger brother is absolutely conscientious and conscientious, and he is a cow and a horse, and he has died with all his might. After experiencing these two technical shocks of Meng Fanqi, Wang Kai has been completely impressed by his technical strength.

In many cases, the algorithm is not a question of whether it can be written, but whether it can be imagined.

In the past few days, the Baidu AI Research Institute has recalled a group of technicians who are on the go, hoping to gradually shift the focus back. But after almost a week, there has been no meaningful progress at all in multi-target detection and tracking.

Although the business of whiteness AI has been thriving recently, and there is a good situation, many internal AI technicians actually have a lot of complaints.

The YOLO algorithm is very strong, but it is not independently developed by Baidu at all, but it is just borrowed the power of Meng Fanqi to do a big landing project.

It is indeed advancing by leaps and bounds in the short term, but if you look at it in the long run, it is not a good thing, and the real accumulation of technology is not enough.

Especially when there are so many projects and clients at once, researchers are greatly distracted, and many people privately complain that they are doing things outside of their duties.

The new demand from the government is a good example, and the lack of time to precipitate the whiteness of the technology is simply at a loss for a while, and it is still necessary to turn around and find Meng Fanqi to do this demand.

Wang Kai felt that instead of going to Meng Fanqi repeatedly like this, it was better to simply go with him.

Based on his perception of Meng Fanqi's technical strength, starting a business to make a lot of money is only a matter of how much money to make, and there is no possibility of failure at all.

After listening to Wang Kai's series of inside information, Meng Fanqi can actually quite understand that the ideas and decisions of the management are indeed not very friendly to the bottom employees.

Because they don't, and can't really consider the problem from the perspective of low-level employees, to put it bluntly, others don't care about the idea of wage earners.

"I really don't have the idea of starting a business in the short term, maybe there will be plans in two years." Meng Fanqi's participation in the form of technology shares alone is enough to earn enough wealth, supplemented by some investment, which will be much easier than starting a business.

If he really wants to run his own company, Meng Fanqi currently has only one plan, that is, two years later, at the end of 15, he and Ilya will leave Google to run OpenAI together.

But even he didn't think about it, because in the planning, 15 years ago, Meng Fanqi would publish the core technologies needed by ChatGPT, such as attention mechanism and generative pre-trained Transformer.

At that time, maybe it's good to invest in OpenAI with technology and capital, and be a technical consultant, and there is no need to start a business and work on your own.

"Let's talk about it then, if I make any moves, I won't hide it, remember to contact me when you see it."

Someone identified himself so early, and Meng Fanqi was willing to let him join the team, after all, he was also an engineer who did AI in Baidu, and there was no problem at the level.

It's just that I really don't have this need in the short term, and even if I do, I'm not embarrassed to poach the people sent by Mr. Li to dock directly.

At the end of the brief call, Meng began to retrieve the Hungarian algorithm and the Kuhn-Mankels (KM) algorithm, both of which are actually old methods.

For example, the Hungarian algorithm, proposed in 1955, is a combinatorial optimization algorithm that solves distribution problems in polynomial time.

The SORT multi-target tracking method only uses these two and the YOLO proposed by Meng Fanqi, and does not require many changes to the three themselves.

The YOLO algorithm is used as a feature extractor to extract all vehicle parts.

The Hungarian or KM algorithm maximizes the matching of several targets between adjacent frames, and the Kalman filter makes some corrections to the results according to the prediction of the target trajectory.

Especially in the government scenario, where the trajectory of the vehicle is relatively regular and simple, the effect of this method will be quite excellent and stable.

In addition, in adjacent frames, Meng Fanqi also introduced the IOU distance, that is, the area of the intersection between the two detection frames, and used it as the weight of the bipartite graph matching.

At the same time, it can also calculate the speed of the vehicle according to the distance between frames, and reasonably design the threshold of relevant parameters according to the speed range, which can be said to kill two birds with one stone.

On top of this, there is an improved version of DeepSORT, which is inspired by pedestrian recognition and compares the distance between two hidden layer parameter vectors to determine whether they are the same person.

However, Meng Fanqi is not so thoughtful, to achieve so much, it is enough to meet the basic needs. This part continues to improve the content, and he is only going to write down an idea here to give whiteness.

It's still more cost-effective to spend time on Google.