Silicon Valley Chinese Entrepreneurship Team Orbeus: die-knock image recognition

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

I can always hear Orbeus's name in the Chinese circle of the Silicon Valley Bay area of the United States (whether it's with investment or with entrepreneurship). Anyway, it feels like purple: in Silicon Valley. If you don't even know orbeus, you'll be dirt. It is a big tech start-up, a variety of award-winning programs, and has been enrolled in a number of star incubators, including YC, focusing on a field called ' Computer Vision ' (aka CV).

So, in the end how a few bull cows, what a bullish thing to do in the Silicon Valley to create a circle so eat? Curiosity drove me to the team's resume:

All two co-founders from Boston University (Boston University), who hold the company's core technology, suspended their original doctoral program for Orbeus and rejected full-time job offer (including Google).

The incumbent CEO graduated from Bently University (Bentley University) and gave up his original analyst job for Orbeus.

But, put the big company's full-time offer not, the annual salary 100,000 knives of the comfortable days, but the sunlit to the green card does not go ... 6 of these bear children are quite daring to play.

Recently, I visited Orbeus's office in the Silicon Valley Mountain View City. This mobile house (moving Villa) is a 6-7-person entrepreneurial world, but also a collective dormitory. A few open large tables, several computers, a few books, the kitchen and the living room completely without transition.

Let's talk about the Orbeus.

This start-up company provides cloud based image recognition solutions, and the core product ' Rekognition API platform ' identifies objects such as faces, scenes, landmarks, objects, and other concepts.

What is the recognition method?

To be blunt, you give a photo, rekognition technology can only according to pixel information, on the analysis of this image is what: is a cat, is a male or female, is a hamburger, is the beach, is the party, is crying, is happy, is Tiananmen, or professor ... In short, this technique can make a fairly plausible annotation of the semantics of the image.

At present, Orbeus's set of in-depth learning based image recognition system can identify more than 2000 kinds of objects, scenes and concepts, can also be used to analyze video. Orbeus's small partners also made an iOS display application, to give everyone a few screenshots meaning Bai.

It can be seen that the relevant concepts and accuracy identified: such as sports car (63%), Racing (57%) and so on.

PK Baidu Physical Translation

Children's shoes have no impression, 1 months ago, Baidu translation app joined the ' real translation ' function, the idea of cute knot, but the accuracy is not flattering, the result was a flurry of media friends molested shelling.

In fact, rekognition this show app a bit like the meaning, but "we are more accurate than Baidu!" before the upload of personal photos was the mother translated into Black Poodle CEO Liyi proudly told me; And their technology is in house, Google is the same, the technology to do a good job is suppressed for their own use, Orbeus want to do AWS as a technology provider, we provide the API, the image recognition technology embedded in any can let it glow of the field, social networks, E-commerce, marketing , picture video collation and so on. "

Founder Liu Tianjiang told me, "To do this demo application is actually a good idea, the image recognition can be used in many areas of hard to imagine, orbeus of course, we look forward to working with the developers of various fields to explore these possibilities." ”

Over the past year, Orbeus has worked with Panasonic, AMD and other vendors to provide technical support to a number of app developers, including Facebook application The Hottie Finder, and has launched a face-recognition SDK for Google Glass developers.

In Orbeus's view, what is the application of image recognition technology in the short term?

1 If you today in the street to see a passer-by's overcoat, Gray often heartbeat, run to the Internet search, you need to enter text to tell the computer: To find Dongdong is a piece of plaid coat, thousand birds lattice pattern, a large, double row buckle ... If you can use image recognition technology, these will be saved. If you take a picture and upload it, an app can tell you the source directly and recommend something similar.

2 Imagine you wearing a Google Glass in India, eyes to see the moment, the relevant information will be directly ejected, tell you this is the Taj Mahal, historical background and so on immediately, this is the so-called land mark recognition, very favored by Liu Tianjiang.

3 with the advent of wearable equipment, video/photo shoots will be more and more, perhaps in the future, ' photo ' will no longer be a conscious behavior. Photos are more and more, can not be sorted out, cough, so, this is the Smart album debut time! With the help of image recognition, you can enter the key words ' wife, bikini, beach ' and so on in the vast picture, the smart photo album will help you automatically organize the pictures that conform to the description. For example, you are called the Beast fans, see "star you" when only want to pick his clips, with image recognition, these are small case.

Finally is the start-up chicken soup time!

3 Experiences of Orbeus venture so far

1 do not believe in the famous school complex. Orbeus, who came out of Boston (where the Harvard MIT is located) did not have a particularly eye-catching background. They ' debut ' that the VC eyes only see the elite graduates, want to find some famous schools come over ' stained double-dipping ', later found that the necessity is not so absolute. Put aside those imaginary head of the brain, they still have the opportunity to rely on strength rather than elite ring to impress investors. The same law can also be applied to ' recruit na-yin ', they gradually learn, do not put labels on the other hand, a good education is commendable, but again praise is just a stepping stone. In the end, it was a mule and a horse that had to be pulled out.

2 Do not come out of the Chinese circle. In the Silicon Valley, the Chinese create the circle has a small circle of their own, head down on those several people, everyone friendly familiar, mutual help. But, the group is not well integrated into the Silicon Valley mainstream. Orbeus first came through America's top incubator (Excelerate Labs), which earned a lot of exposure to mainstream media in the United States because of its eye-catching performance and investment. In their view, the participation in the incubator draft is certainly not the only way, but it really benefited them, in addition to the U.S. market to understand the orbeus, but also let them understand the American circle of the rules of the game.

3 don't think of yourself as a Google engineer. Even if the technical background also need to understand some business, don't think with ' business ' of those things, find a person to do it. Keep in mind that it's not as simple as being an engineer in Google. Liu Tianjiang Frank Orbeus has paid a considerable price for this, his suggestion is that, as the founder of technology, usually should also turn over the book of business, "in fact, it is not difficult to understand, heavy in the conscious accumulation," Liu Tianjiang said.

With regard to ' competition ', Orbeus gives the following advice

Entrepreneurs, don't be afraid to compete, try to see it as a natural thing. You think so, a person want to fry hot market is how difficult thing, your team as long as in the competition in the first echelon will be able to, the remaining people are in for you to do BD (business development).

"Besides, you don't have to beat your opponents, you just need to ' stick ' and wait until 99% of people hang up and you win," said CEO Liyi.

Orbeus is not the first company to enter the field of image recognition, but unconsciously, when a competitor fell or was received, they suddenly look back, found ' Yi? ' Where have all the people gone? How to become the biggest independent technology provider in this field! Hey, this world is sometimes so incredible.