The top international computer vision conference in December 2013 was actually a little longer.

The conference program is divided into two days of seminars, 1-2, then the main conference is 4 days, 3-6, and finally two days of seminars on specific topics, 7-8.

The IMAGENET-2013 event won by Meng Fanqi is scheduled to be held on December 7, 2013.

Therefore, strictly speaking, Meng Fanqi does not need to attend the main part of the conference.

But he wasn't going to go too late, and the group booked a ticket on the 4th.

For the first three days of the beginning of December, nothing special.

Meng Fanqi went to Baidu's Yanjing headquarters again, this time because of Baidu's propaganda needs.

The Deep Learning Research Institute on Li Yanhong's side acted very quickly, and soon studied the key points of Meng Fanqi's method, and began to try to reproduce it on his own data.

Wang Kai, a whiteness researcher who is responsible for docking with Meng Fanqi, frequently visited Meng Fanqi near Yenching Electric Power University during this time, and almost thoroughly understood all kinds of relevant knowledge.

Meng Fanqi still prefers to deal with more pure people like Wang Kai, although it stands to reason that Wang Kai's contact frequency is a little higher, and the questions he asks sometimes exceed the reasonable range, but his methods are quite exquisite.

Since he didn't have anything particularly important, Meng Fanqi didn't care too much.

Shortly after the last signing of the technical contract, under the active promotion of Wang Kai, the algorithm has been mastered in terms of whiteness.

Robin Li immediately made a decision, prepared materials, and wanted to make a promotional video with the theme of "real-time detection can subvert the world".

To put it plainly, it is to show the use scenarios of Meng Fanqi's YOLO algorithm in all walks of life.

Its core is to highlight the characteristics of several aspects, wide use, fast and efficient.

Under the boss's order, the algorithm researchers on the visual side of the research institute sat in a row and began brainstorming.

They need to think about the requirements within the scope of these scenarios, collect images, and run detection algorithms on them.

These materials are then given to the following marketing department, which is responsible for designing copywriting, editing videos and other operations.

"First of all, it must be in terms of autonomous driving, road signs, vehicles and pedestrians, and the detection of these aspects is not only significant, but also the most in line with the characteristics of 'real-time'. The need for real-time in autonomous driving is the most indispensable. Wang Haifeng made a start.

"Autonomous driving is not only in line with the theme, but also has a very large market, one of the four basic needs of clothing, food, housing and transportation, and its impact is also the most exaggerated. Moreover, its implementation is also very intelligent, with a sense of the third industrial revolution. A researcher nodded that the scene of autonomous driving is the most in line with the theme and the most shocking.

"There is also industrial inspection, which may be defect detection, inspection of defective and defective products, or statistics and counting, such as counting the quantity of a certain material and counting." Wang Kai believes that different industries have their own needs, and this aspect will be a great business opportunity.

"However, in industrial inspection, the real-time attribute is not so strong, and many operations can get results in a few seconds."

"But we can emphasize efficiency, we can make a slightly shorter video, on the left is the worker's own point detection, in the middle is a very slow detection algorithm, and on the right is our new algorithm. Speed up the video and compare the number of results they have. ”

"That's right, at the end of the year, give a percentage of the result, like a 160 percent efficiency increase or something."

"There's also a big market in terms of medical imaging, and that's something we definitely need to mention. However, compared with autonomous driving and industry, the classification and detection of medical images, the concept of real-time is even weaker, and there is basically no situation that needs to be identified in real time. Some researchers have raised this difficulty.

"And medical responsibility and moral issues, we can only say that it is an auxiliary method, and we cannot promote alternative doctors, which needs to be paid attention to." Yu Kai passed by and listened for a while, and reminded loudly.

"And security detection...."

"Intelligent annotation of key periods of long-term recording...."

"Food Testing ....."

"Animal testing...."

"Face Detection...."

"Pedestrian Detection ....."

The crowd worked together, whether it was really a chance to land immediately, or it was completely a pie, and it felt like basically no one would pay for the scene in three or five years, all of them would be put forward and counted first.

"I'll get here today, and within two days, at least 20 picture detection results will be prepared in each direction, and there will be five short videos measured by the camera as material to be given."

Yu Kai finally concluded, "The picture materials will all be ready to be given today, and the short video can be a little slower." ”

In the short term, the primary task of the Baidu Research Institute is to assist the marketing and publicity departments to promote the versatility, high performance, and technological disruption of the YOLO algorithm.

According to Li Yanhong's idea, the speed of this promotional video must be fast.

The speed of AI technology is so fast that it would be a little more shocking a day earlier.

Medium- and long-term plans should be based on real feedback from the market, and only after liaising and talking with major industry leaders can we formulate and plan.

After Meng Fanqi came to the headquarters of Baidu this time, it was not that the three giants were responsible for docking, nor did he go to the Deep Learning Research Institute, but he was directly pulled to the marketing department.

As soon as I entered the door, I was pressed directly onto the chair by the two sisters, and I started to put on makeup in front of the mirror.

"Engage in such a big battle?" Meng Fanqi had an experience of hosting in the hospital when he was a freshman, and as a boy, he was also exposed to makeup for the first time, so he probably knew what to do.

It's just that I didn't expect it to be such an exaggeration, and there are three professionals just by smearing it on my face.

This made Meng Fanqi really flattered.

"We still have a lot of things to shoot today." Li Yanhong lacked skills, and this time he sent his personal secretary Liu to accompany him. "I heard Mr. Li say that you are going to attend the International Computer Vision Conference in a few days, and before you leave, these materials must be filmed well, and they will be ready to be sent immediately."

"Moving so fast?" Meng Fanqi knew that Robin Li had always wanted to make practical achievements in AI, and his YOLO algorithm gave him what he needed most, but Rao was so, and Robin Li's movements seemed too fast, and it was only less than a week, and the promotional video was about to be released?

"Let's take a formal bust photo first, this will definitely be used at the press conference. After a while, I will shoot a few materials to explain algorithms, write code, and do testing experiments. Finally, when Mr. Li comes back, I will shoot a few more materials for your exchange and discussion. That's pretty much enough. ”