Rewrite the Technological Landscape

"Mr. Meng." After the meeting, Zhuo Liangcai and Cao Shuang came to Meng Qian.

"What's the matter?" Meng Qian noticed that their expressions were a bit strange and motioned for them to take their seats.

"We..." Zhuo Liangcai glanced at Cao Shuang, "We want to start a business."

Meng Qian took a sip of tea calmly, "What entrepreneurial project?"

"Future search."

"Future search?" Meng Qian was a little curious. "The next-generation search engine for a little search?"

"No." Zhuo Liangcai, "It's a creation.. I still dare not say whether it is a creation, but it should be a subversion."

"Listen specifically."

"As you just said, the key to using first principles to subvert is to see the nature of the business, then what is the nature of search.

In our view, the essence of search is to ask questions.

When people ask questions, obviously, it is when we encounter problems.

Then there is a problem here, that is, timeliness. Today's search engines are designed based on computers in the past. The original logic of this design has a limitation, that is, it is not removable, because before the popularization of laptops, computers can Said to be an immovable product.

So when we encounter problems, can computers appear by our side in time? Recalling carefully, we will find that in most cases, the computer will not be by our side. For example, if I encounter problems when I go out, I cannot search.

Until the emergence of smart phones, the emergence of search engines on mobile phones has solved this problem, but mobile phones as mobile devices are still using search engines based on immovable computer logic. This makes us start thinking about this problem. Is there a problem? "

Meng Qian put down the tea cup unconsciously, Zhuo Liangcai's expression attracted him, "Let me give the simplest example, when you take a bus, you see someone holding a Hongmeng 3 in his hand, and Hongmeng 3's The appearance attracts you and makes you want to know about it.

There are two situations at this time. The first is that you know that it is Hongmeng 3, you can take out your phone and open a little search to search for Hongmeng 3, and the relevant information will come out.

Another situation is that you don’t know what phone it is, and you want to know what it is, what should you do? You can only go up and ask, but what if you are embarrassed to ask?

At this time, the existing search engines seem to be a little weak in solving the problem you are encountering at the moment.

So we began to consider whether there is any other way to solve this problem, until we think of a possibility, if we can let the mobile phone recognize our curious target, and the mobile phone will feedback the result to us, that is, through Is it possible for the mobile phone to use its own identification to complete the search purpose?

If possible, can we name it, visual search.

The user only needs to turn on the camera and aim at the target to obtain relevant information. "

"This can only be regarded as visual recognition." Meng Qian frowned slightly, slightly disappointed. "As long as the database is built and the algorithm is designed, any mobile phone can have this function in a few years."

"If it only displays the basic information of the object being photographed, it is really just a visual recognition, and our company will soon overcome this technology.

But if you add user behavior understanding, it's not just recognition, it can then evolve into a search function.

For example, when we travel, as a mobile phone, it has your location and a series of data about you, such as time, recent itinerary, travel mode, because with the help of mobile phones, the information that the algorithm can obtain is beyond the picture or video itself. Yes, the algorithm can use these data to guess what the user most wants to search for and sort the search results. A series of feedback results like a one-point search will appear instead of just the basic information of an object.

Users can find what they want in this series of feedback results. Of course, visual search can be combined with voice and text input. When the algorithm makes a mistake, users can refine the content they want to search through voice and text.

This vision-based search application will have many scenarios that can be expected, such as learning, such as social interaction, and outdoor commercialization. "

"To achieve visual recognition and user understanding at the same time, the development of this algorithm is not easy." Meng Qian pointed out the difficulty directly. After all, this is a technology that has not been implemented in 2019. "The most basic visual recognition alone has big problems. .

Your visual recognition is not the same as the visual recognition we are conquering. The current development of computer vision technology is based on point clouds, that is, by scanning objects in 3D, building point clouds, and then using point clouds to train machine algorithms. For the machine to be able to recognize this object, you should be very clear.

However, this counting principle is more suitable for indoors. It is more effective in identifying objects and faces. When outdoors, the recognition ability becomes very poor. Most commonly, as the light changes, the color and intensity of the reflection of the building will change. The shadow will change. In addition, pedestrians and parked vehicles around the building will also affect the machine's recognition of it. How are you going to solve this problem? "

"We used another algorithmic logic." Zhuo Liangcai put the prepared tablet in front of Meng Qian, "Our recognition is based on geometry and labeling.

First of all, we must build a huge database, and as users upload more and more information, our database will become larger and larger.

Secondly, we will perform characteristic geometric processing on outdoor objects, such as the shape of a mountain, the shape of a tower, these very characteristic shapes, so as to reduce the dependence on light and other information that changes greatly.

Then we will label the information in the database. This time is like the positioning I mentioned earlier. Under the premise of knowing where the user took the picture, we can quickly find the location through the label, and then combine the geometric information. Good to confirm the goal. "

"When you first started thinking about this technology, what was the source of inspiration?" Meng Qian asked seriously, "Isn't it just considering the nature of search? Your two research directions have never been search."

Zhuo Liangcai smiled with Cao Shuang and said, "One day we saw a strange-looking stone in Xixi Wetland, and then we stared at that stone for a long time, but our brains were empty.

Curiosity is actually human nature. Sometimes, for example, when we see something we haven't seen before, we are very curious, but let us use words to express what we want to search for and what we want to understand. In fact, we are not very clear.

At this time, visual search will become very valuable. When you are curious, you only need to take out your phone and take a shot, and a lot of information will appear. There may be something you want to know about it.

This is indeed our original intention to do visual search. We found that most people were full of curiosity about the world when they were young. Flowers on the roadside and ants on the ground can make us have a desire to explore, but as we As he grew up, this curiosity became weaker and weaker.

Maybe visual search can re-inspire the curiosity of more adults. "

"To be honest, although you have a certain direction, in my opinion, the technical difficulty may be higher than your estimate. But your ideas really touched me. Are you sure you are ready? not easy."

"We have not underestimated the difficulty of this technology, and we have been prepared to continue to die, for three, five or even ten years, we think this is a promising technology, it is meaningful and has commercial value. ."

"Commercial value? This should be the question Mr. Shao should ask. Now that you mentioned it, what is your business model? How to make a profit? I didn't see it at all." Meng Qian asked along the way.

"If visual search can finally be realized, then the profitability of search and Google will be our future profit model."

"It takes too long to wait~www.ltnovel.com~ before that?"

"In this process, the technology of outdoor visual recognition alone is enough to provide us with opportunities for business cooperation. Now many companies want to obtain user outdoor information for big data analysis, then our technology can be used. Up."

"So you estimate how long the initial investment will take, and when will you see a return?"

"Three years, one hundred million."

It can be seen that Zhuo Liangcai and Cao Shuang are taking this matter very seriously, and Meng Qian is really looking forward to this technology, "Okay, go to Mr. Shao with your business plan.

Leave the business judgment to him. As for technical issues, I am willing to trust you. "

Zhuo Liangcai and Cao Shuang were overjoyed and brought their business plans to prepare for their new journey.

Meng Qian looked at the backs of the two leaving and looked forward to it silently, "I hope more people can surprise me."

...