Iron Man 2 small Luobo·tetangni fingers to manipulate virtual program
In ' iron man 2 ', young Robert Downey gestures can make the computer screen opens, virtual programs pop open hanging in the air, can also play games, bunched up into a ball to virtual box program.
This image, involved two terms, a sense of recognition and human-computer interaction. In commercial applications, we have not reached the point of comfortable cool in the movies. But in the domestic science and technology circle, done through the identification of breakthrough progress in the human-computer interaction, and application there can stay.
Front technology team in a building in Chaoyang District of Beijing. Office space, but the pattern more open, Sun is shining. They are entrepreneurial team of gesture recognition using machine vision technology, the company set up for 4 years, matrix partners China and actually Fund has led the angels and Pre-A rounds.
From a research and development company was founded five years, Feng team when only a gesture recognition products, secondly, this mentality is what is most difficult to imagine their practice.
"F4 front" was born from the feelings of life-changing technology
In 2010, the ustc graduate Liu Jin Su (Su) and Liu Zhe, Dang Jianxun, zhangshuo three people get together, talk about how to do something. Somatosensory interaction they intend to apply machine vision technology, this is a developing trend in the application of science and technology. So the four generation make 500,000, was set up in July 2011 front technology group "F4 front" was born.
Among them, Liu Jin Su originally conceived of the project and the practitioner.
From 2006 to 2008, Liu Jin Su in science and reading blogs and visiting scholar at Carnegie Mellon University during the respectively won the RoboCup soccer simulation-d Championship and simulated SPL title. Machine vision technology plays an important role in both events.
Through the game, Liu Jin Su machine vision in the next machine automation and human-computer interaction fields have a broad space for development. Liu Jin Su after heart shock give birth to an idea, he was going to do it--the application of machine vision technology into people's daily lives.
"Machine vision technology services for people living really" had a front technology founded the beginner's mind.
If you are not a technology entrepreneur, or technology backgrounds, this sentence can be easily considered to be upper and lower lips move a muscle thing. Only when they are subjected to temptation, to know the importance of ideal composition. They said that technology didn't do, things not to do beautiful, unwilling.
Teams of four people, developed to 30 people, "Jing-Shen" separated, bulkhead is five years, I don't know where the time has gone?
5 years six generations of the iteration history stories
From development to the establishment of the company's five-year fumble, when the front team breakthrough in gesture recognition for machine vision applications.
Gesture recognition is a kind of human-computer interaction, people across an empty gesture, control systems. The core of this technology is that through the fusion of two cameras to capture the hand of spatial information, complete a command input and transfer, the ultimate goal is learned through the gestures used on the operating system the intention.
First generation of pre-research December 2010
When back in front before the establishment of science and technology, December 2010, Liu Jin Su envisaged first to the first generation of advanced machines. It is made of wood and black plastic box with a concept, although quite rough, but it verifies the feasibility of machine vision applications.
And the back of the finished product than this large value is--it was the start of a start, like the Wright brothers ' plane, like a Turing machine.
2011 12 second generation micro engineering
In December 2011, the team established five months after they made a second generation machine. Although a lot smaller than the first generation of big, but still rough, connected three cameras two pieces of iron are removed from your children's toys. To gather information, when the machine is configured with three cameras, each camera 24 IR lamps, power consumption is very large, requires a 220V voltage support. Collection of pictures to be transmitted to the PC side, algorithms, cumbersome, and operation pressure is relatively large. However, "I have a mold, can take out the said thing," CEO Liu Zhe said.
So they brought the second generation machine to Lenovo. Lenovo recommends them, first of all, keep the power consumption down and complete image processing integrated in the product and get rid of its dependence on PC operational. Most importantly, suggesting more "intentional." Not only does this provide them with improved, and let them know to terminal manufacturers, especially consumer electronics demand for human-machine interaction technology.
July 2012 third-generation micro-engineering machine
And so they continue to improve. In July 2012, the third generation of micro engineering machine was born. Micro machine marked great changes have taken place in the structure, models also structured a lot.
December 2013 fourth generation prototypes
In December 2013, Feng real product prototype finally was born, this is when the front team to launch the fourth generation machines. Compared with previous ones, prototype has a beauty at all. They will drop to 5 v, algorithm for image processing is done in small boxes, the transmission speed for 500M/s.
Liu Zhe said, "other teams, are based on PC or host platform for image processing, we have set aside their heavy-weight framework, validation of mobile in the future possibilities in the technology product form. "After you have a prototype, they've got 6 million Yuan into the matrix partners China led an angel round of funding.
In 2012, there was an episode. There is a "big brother" also after the project, wants to spend 5 million acquisition. 5 million Yuan was initially invested 10 times, 100,000 in the vote, for example, two years to $ 1 million. When you set up the front panel, the four founders graduated, some just quit my job got married, others are studying abroad. 5 million equivalent to get 10 times return on their first pot of gold.
In view of the development prospects after changing hands, whether you can make the project continued to develop in depth, four founders are doubts. "That we do have a certain temptation", co-founder, zhangshuo recalled, "but the team did not appear dilemma, even useless thoughts to think about. Just tired and occasionally joked about this matter. "
Finally, they certainly didn't sell.
"The feeling that something needs to be done, not so beautiful. "Liu Zhe bulkhead reflections said with a smile.
2014 fifth generation vidoo
Get financing, in 2014, the team continued to improve products. Solves supply chain and production grinding problems, sample shows the transition. The end of 2014, the fifth generation was the first formal front technology products to market vidoo, official price 399 Yuan. At that time, vidoo also participated in the naming of all the chips, finally all raised amounts to 620,000 yuan.
620,000 may not for the other team is nothing, but for five years, one year a prototype iterations, through three core algorithm upgrade "product" can be improved, which is when the front team is an inspiration.
At this time, Feng also finished the Pre-A round of funding from matrix partners China and serious investment.
December 2015 the sixth generation the second generation Vidoo
Blink of an eye by the year 2015. Second generation Vidoo is expected to be available in September this year. Second generation Vidoo looks more aesthetically pleasing, and new technology such as increasing MID and trajectory of the hand enter, redesigned lens will provide 120 degree viewing angle range.
"Hands in the air, and in writing on the paper is not the same. Writing hand because the air trajectories is constant, so the system can recognize your gestures, this algorithm is difficult ", Liu Zhe stressed.
To solve this problem, however, is essential, so they designed the air path identification algorithms. At present, this algorithm can identify 10 Arabic numerals, 26 letters of the alphabet and five-pointed star, triangles and other simple graphics.
Difficulties in somatosensory interaction, market size and when the winger's future development
On application of gesture recognition, Google, Microsoft, and recently acquired by Oculus Israel company Pebbles are doing, including Lenovo's 2015 Techworld Conference, has any body on this show. The system operation is relatively simple, only need the palm facing the screen, players from top to bottom and move around the Palm, lead villain in the game to the specified location. However, the experience of "small white" users need to adapt to a child.
CEO Liu Zhe (right) and co-founder of zhangshuo (left)
"Although there is a large company in somatosensory interaction, but they usually only with this technology as a product feature or component technology to meet the needs of the product, will stagnate, so as to be of any real technical level ' well done '. "
How to understand this "good"?
"Especially in somatosensory technology and precision delay", said Liu Zhe, they used two or three error control in 0.01 mm accuracy of gesture recognition, the delay time in 10 milliseconds.
The other hand, the computer screen when played, sensitive operation, but for users with small white, use habits is the first hurdle. Users in the real scene interacting with two-dimensional images on the screen, will cause, hands, brain and eye of disharmony. This will increase the user's extra burdens.
In order to overcome this difficulty, when the front team to design a UI interface solutions. Two-dimensional icons on the screen does not change screen interface designed three dimensional skeletal hand gestures. Hand made the pick action above the vidoo, will appear on the screen hand gesture, to grab the target icon to a file to the Recycle Bin, and so on. This technology, eliminating the user thinking spend trouble and identification operations.
Many scenes of daily life using somatosensory interaction. Receptor interaction level of technological development and cost constraints, Feng, team with VR and gaming as a breakthrough, TV, smart homes, cars and even medical scenarios. They have been in the same front loading of TV makers and car manufacturers to conduct exchanges and cooperation.
However, vidoo just at the front of a tangible product. Gesture recognition is the core technology of machine vision algorithms. On current science and technology and the development trend of the evolution of life, universal application of somatosensory interaction will be the future. So, soon after when winger's goal is, and tangible products of small, gesture recognition sensor, integrated into the device boards, so that they became interactive solutions provider.
Somatosensory interaction how big is the market?
--"Imagine space is big," said Liu Zhe. Wit founder how to imagine everything Internet