Page 334
——This sounds really reasonable and well-founded.
Zhou Yunlong: “…”
Zhou Yunlong: "So, how can Li Yuhua be so strong in building structures?"
Wu An smiled again and said, "I taught her well, didn't I?"
Zhou Yunlong: “…”
Zhou Yulong didn't believe Wu An's lies.
But one has to admit that “that Li Yuhua” is really talented.
His level is at least not worse than Wu An’s.
The only explanation is that she was indeed taught by Wu An.
But why would a student from an art college want to get involved in artificial intelligence such as computer vision?
Can't understand.
In short, Zhou Yunlong and Wang Zixia did not feel that their work was hindered by Wu An’s lack of participation. Instead, they completed the compilation quite smoothly and entered the stage of deep learning training.
The progress is faster than expected.
By the afternoon of the next day, the framework of the deep learning algorithm was completed ahead of schedule.
After using the framework to start deep learning training, everyone can take a rest.
Zhou Yunlong and Wang Zixia took the time to recuperate.
Wu An, who plays Li Yuhua, worked on the code all night and planned to go back to rest.
——Although I feel like I can still hold on, since I have already completed a stage, there is no need to force myself to do so.
Wu An asked Li Yuhua to stay on the scene to help deal with situations that might arise during deep learning, and went to sleep.
Wu An slept until after eight o'clock in the evening.
When I returned to the competition site, I saw "that Wu An" sitting in front of the computer, adjusting the algorithm model based on the results of several rounds of deep learning calculations.
When they saw Wu An, the two looked at each other. While Zhou Yunlong and Wang Zixia were not there, the two of them squeezed together and started adjusting the algorithm model.
The tense competition scene was filled with a sweet atmosphere of love.
After resting enough, Zhou Yunlong and Wang Jiarong, who came to the competition venue from the room later, were suddenly licked to the point of choking.
Ah!
Really!
I've had enough of this couple!
#132 We are actually very strong
Zhou Yunlong and Wang Zixia complained about Li Yuhua and Wu An for a while, and then joined the battle to adjust the algorithm.
First, we checked the initial architecture for deep learning results. The results...were just okay.
The real-time distance of objects identified from two-dimensional images is not as accurate as imagined.
The measured distance constant drift is relatively large.
Using Kalman filtering and high-pass and low-pass filtering algorithms to perform some filtering will also result in deviations in the test results of object distance.
This situation is reflected in the modeling, which is wildly erratic.
But... is it still usable?
Wu An and his team didn’t know what the other groups’ work results were, but they themselves felt that they were not perfect.
But...at least it proves that their deep learning algorithm is correct.
All that remains is to optimize the algorithm as much as possible.
The time that follows is a busy and fulfilling period.
Every time the modeling progress improves a little, they will feel a huge sense of accomplishment.
It is also embarrassing because the new algorithm has a negative effect.
Later in the evening, Li Yuhua made an excuse that she was a little sleepy, left Wu An and the others behind, and went back to rest.
Zhou Yunlong and Wang Zixia looked angry, but seeing that "that Li Yuhua" had no objection, they could only feel indignant in their hearts.
Wu An said to them: "As long as Huahua is here, it means our souls are here too. You guys should keep working hard!"
Zhou Yunlong: “…”
Wang Zixia: “…”
Well, Wu An is right.
"That Li Yuhua" is indeed quite amazing.
It was as if Wu An was there.
Well... since there is no problem, just keep going.
Just like that, at four o'clock in the morning, Wu An and Li Yuhua took over their shifts.
Wu An returned to his own body and started another round of hard work, while Li Yuhua went back to sleep to give her body a rest.
Zhou Yunlong and Wang Zixia could not have Li Yuhua help them "take over" and rest like Wu An did, so late at night, they also put on a fairy-like look.
Seeing that the algorithm had been slightly optimized, Wu An asked them to go back and rest, and come back to modify the algorithm in the next stage.
In this way, Wu An survived until the morning of the third day of the competition.
It was almost nine o'clock when the group came back.
During this period, Wu An came up with new ideas for improvement by observing the deep learning training model.
But Wu An couldn't finish it so quickly by himself, so he assigned it to Zhou Yunlong and Xia Ziwu. Before he knew it, the day was over.
The new algorithm did not bring any improvement to the modeling, which made everyone a little frustrated.
On this day, Li Yuhua secretly went to see the situation of the teams taking part in the same test, but couldn't figure out what was going on.
——It seemed that they were stepping into the same pit that Wu An and his friends had stepped into.
What we have to compete for now is who can get out of that pit.
If I can come up with a more effective solution than others, then by then, maybe there won’t be a solution available.
Late that night, everyone gathered around Wu An's computer, looking at the modeling results that had not shown any improvement, and the atmosphere was somewhat depressing.
"It's different from what I imagined."
"Mr. Wu, what should we do next?"
Zhou Yunlong and Wang Zixia asked Wu Andao together.
Wu An didn't know what to do next.
However, failure is the mother of success, so keep trying!
Wu An stretched himself lazily and said helplessly, "Let's leave it this way for now. It's normal for there to be a gap between theory and practice."
"If there is no other way, everyone should go back and rest."
"I'm just like a headless fly thinking about this, I don't think I'll get any results."
"Tomorrow is the last day, so don't change the plan anymore. Think about how to optimize it."
"Maybe the model we made yesterday was already very complete?"
He said optimistically.
Seeing that Wu An thought it was okay, Zhou Yunlong and Wang Zixia also thought it was okay.
If you can’t do it, just go to sleep!
Maybe there will be a miracle when we wake up tomorrow morning?
Just like that, all members of Wu An's team went back to sleep.
The computers at the competition site were turned on, running programs for deep learning.
——What if we train more, our progress might improve?
Wu An and his friends left Zhou Yunlong at the venue to watch over the game, while they went back to the hotel to take a shower and sleep.
The disappearance of Wu An's team is like giving up hope when tomorrow is the last day of the competition and all the participating teams are scratching their heads and sprinting.
However, when someone noticed the deep learning program running on the computers in Wu An's team's area, a huge sense of urgency came over them.
Because the algorithm that Wu An and others consider to be "not so good" may not be the case in the eyes of others.
In the eyes of other participating teams, Wu An and his team have already implemented the algorithm and are conducting intensive training in deep learning.
However, most of the participating teams were thrown into disarray by this "new question" and have not been able to come up with a reliable algorithm until now.
too strong.
That team is so strong!
……
The next morning, Wu An and Li Yuhua got up and came to the competition venue leisurely.
Wu An didn't sleep well when he went back last night.
Because when Wu An was lying in bed yesterday, he was still thinking about what went wrong with their code and architecture.
After thinking about it for a whole night, I realized that the problem... naturally has no solution.
Later, I fell asleep due to mental fatigue.
Today, he and Li Yuhua exchanged bodies.
When we arrived at the competition venue, we first checked the results of the artificial intelligence training last night.
This sight made Wu An slightly stunned.
After fool-proof deep training, the recognition rate of the program under certain specific conditions has actually improved a lot.