The university entrance exam
It’s always a dress rehearsal before the show
…
The Top scorer has always been the talk of the town after dinner
Year after year, from the imperial examination to the college entrance examination, immutable
Ten years of hard study no one asked, one famous world know
This is indeed the true portrayal of the current college entrance examination system
The top scorer in the college entrance examination was mobbed by reporters
For the vast majority of mediocre students, however, the top scorer may be sour grapes in the distance
No matter when someone will attack the top scorers from different angles
But the work and life of the top scorers after graduation is the hardest hit area
The famous female game anchor is Shi Yue, who was the top scorer in Inner Mongolia science in 2006
In 2011, the Central Academy of Education and Science conducted a survey about the top scorers in the college entrance examination
The results were published in the Shanghai Education Journal, it said
“We surveyed 3,300 top scorers since the resumption of the gaokao, and none of them became industry leaders.”
We don’t know how reliable this report is
But since 2000, at least one top scorer has become an industry leader
He is He Kaiming, who was also the top scorer in the 2003 Guangdong Science College entrance Examination
He Kaiming
In 2009, IEEE International Conference on Computer Vision and Pattern Recognition
He Kaiming with a picture dark channel prior fog algorithm paper technology surprised four
Of the 1,450 papers submitted to the conference that year, only 393 were included
And He Kaiming’s paper was the only best paper that year
For the first time, an entirely Chinese team won this award!
Hok Ming (right)
Incredibly, this is Kaiming He’s first paper
At the time, he was an intern at Microsoft Research Asia
…
He Kaiming grew up in Guangzhou, living in the historic Liwan District
His parents both work in a management position in an enterprise
As an only child, He Kaiming grew up under the careful care of his parents
Liwan District’s ancient litchi Bay
Perhaps it was influenced by the many educational relatives in the family
Encouraged by his father, young Xiao He entered the children’s Palace early to study
At the children’s Palace, He kaiming studied painting, and he really enjoyed it
Often an afternoon or even most of the day was spent in front of the sketch table
He Kaiming has always been calm and patient character may be an important factor in his success
According to his head teacher, He kaiming’s grades are very stable and his foundation is very solid
He also won the first prize in national physics competition and provincial chemistry competition when he studied in Zhixin Middle School
So he was recommended to Tsinghua University in May before the gaokao
But he still took the gaokao, perhaps because he had been recommended to Tsinghua
During the college entrance examination, I was relaxed and had no burden, and played an excellent level
He won the top prize in science in Guangdong Province, and 2003 is printed on the calendar of that year
With he Kaiming in the same year, there are pea pod CEO Wang Junyu
I entered Tsinghua University as the no.1 scorer
He Kaiming gave up the original recommendation of mechanical engineering and automation professional
Chose the more challenging basic science class
The basic Science class was set up by Tsinghua University in 1998 to train talents in mathematics, physics and other basic sciences.
He Kaiming’s study in Tsinghua University is still steady and hard
He won the scholarship three years in a row between 2003 and 2005
And a brighter future is at hand
In 2007, before graduating, He joined Microsoft Research Asia (MSRA) as an intern
Microsoft’s large basic research facility in Asia is made up of 90% Chinese
Out of interest, I took computer graphics and image courses in university
After joining, he also chose to join the visual computing group
Microsoft Research Asia
However, I majored in basic science of physics and mathematics
That knowledge is simply not enough for research work at MSRA
When I read an article, I often don’t know which methods everyone is using and which authors are contributing. Everything I see is new to me.
In the first year of his internship, He Kaiming was under the guidance of his tutor Sun Jian
I did a few different projects, but without exception, I didn’t succeed
Jian sun, who was MSRA’s chief researcher, has now moved to Face++ at a high price
While the facts are discouraging, He says he learned a lot
These failed projects led him to try many different directions
This is what tutor Sun Jian has been encouraging interns to do
It was this relaxed and free research atmosphere that laid a solid foundation for his future research
He Kaiming
Despite being universally regarded as a “genius”, Cumming was never one of those bookworms who only read books
It’s hard to believe that Kaiming He’s earliest work on fog removal came from computer games
Perhaps it has something to do with his schooling in Beijing
Fog in the game Silent Hill
Once, while playing the game, he observed that there was a lot of fog in the game
These scenes are obviously fictional and do not exist, but can also achieve a realistic fog effect
Further analysis confirmed that the generated images had very different statistical patterns from nature photographs
So he was convinced that human vision must have an effective mechanism for perceiving the presence of fog
This mechanism can be used to remove fog from an image and is definitely superior to existing methods
Smog in Beijing
So he began to study the equations of fog in the image
Gradually he realized that the equations of fog were familiar, as if he had seen them before
It turns out that it is very similar to the Matting equation I studied earlier
This brings great convenience to the research of fog removal algorithm
In this way, he only needs to estimate the fog concentration in the area of the image
Photos with fog and their dark channels
Previously, the most effective way to remove fog was by detecting the darkest objects in the picture
And then estimate the fog concentration, so as to remove the globally uniform fog
But in many cases, the fog in the image is not uniform
This is also the difficulty of image defogging, but He Kaiming wants to challenge
Due to the inhomogeneity of the fog, Kamin naturally wanted to use local dark object detection to deal with it
Surprisingly, a number of experiments have shown that this simple idea works
This is to let he Kaiming hit the right method, but the difficulties have just begun
Although the method is simple and efficient, the experimental results are also very beautiful
But the tutor Sun Jian was not eager to let The method of writing an article published
Instead, they repeatedly ask for the essential reasons behind the success of the method
Dark channels for fog-free images
So He Kaiming again with the problem began to study
Since local dark object detection defogging is successful
Then it means that there are dark objects in every part of the image after defogging
This means that there is a corresponding statistical pattern in the image without fog
To this end, He kaiming counted more than 5,000 images and verified the rule of the conjecture
Know how to do from first understand why again, he Kaiming is finally completed a complete study
The efficient defogging algorithm theory of Dark Channel Prior is proposed
According to the image processed by He Kaiming algorithm, the top is the original image and the bottom is the processed image
But in the process of thesis writing, He Kaiming often quarrel with himself
Ask yourself repeatedly if your point of view is correct and reasonable
Very not easy to convince himself, but also by Sun Jian questioned again
In this cycle of struggle, this excellent paper is finally written
It is such demanding requirements, can be proud of the results
He Kaiming’s paper received the highest score from the three reviewers
Meanwhile, Kamin’s speech in Miami was considered by the audience to be the most interesting speech at CVPR that year
This Best Paper Award is well deserved
It is the first time for a Chinese to win this award in CVPR
In view of he Kaiming’s steady style, this best paper award is no accident
After graduation, He entered the Chinese University of Hong Kong for postgraduate study
During this period, he continued to participate in related research in Microsoft Research Asia
In recent years, He has been involved in computer vision recognition research at Microsoft Research Asia
His team in the ImageNet image recognition contest held at the end of 2015
“Image recognition depth difference residual learning” system outshines others
Beat Google, Intel, Qualcomm, won the first place
Once again, an all-Chinese team is on top of the world
The following year, the team’s paper won the Best Paper award at the 2016 CVPR Conference
And the first author of the article is still our top scorer in the college entrance examination He Kaiming
He has also won the CVPR Best Paper Award twice as the first author!
In August, Mr. He left Microsoft Research Asia, where he had been blessed
Joined Facebook’s AI Research team (FAIR)
There is reason to believe that his talent will not be a “short-lived injury”
The college entrance examination is always just a drill
To get good grades is to be proud and smug
I’d love to tell the media about all my eccentricities
The college entrance examination into a metaphysical atmosphere of worship Buddha
It’s hard to believe they can do anything
Pray for the college entrance examination
I don’t know, a lot of people have asked that question, but I’ve never been quite sure, maybe it’s just a good sense
This is he Kaiming became the number one in the college entrance examination to answer reporters about learning secrets of the answer
His time in Microsoft Research Asia has always been about learning
Read a lot of papers, constantly improve English, repeatedly review their own results
It is this humility that has made him what he is today
Cattle!
Actual combat, never with scores on heroes