Edited by Guojin Securities: "the feast of artificial intelligence is coming."
Computer vision is the most concentrated field of AI enterprises at home and abroad.Computer vision technology companies account for about 40% of global AI enterprises and 46% in China.
Computer vision is the track with the richest application scenarios and the greatest commercial value in the field of AI.. At present,There are four main types of data processed by AI technology: text, voice, image and video.From the perspective of information dimension, the dimension from text to video is increasing progressively, the text has the least information dimension and contains the least amount of information, and the video has the most information dimension and contains the largest amount of information. A higher dimension of information and a larger amount of data, soComputer vision with image and video as the main processing object has richer application scene and commercial value than other AI technologies which mainly deal with text or voice.
Take China as an exampleCurrently, the total valuation of the four head start-ups in the computer vision industry (Shangtang, Kuangshi, Yitu and Yuncong) has exceeded 100 billion yuan, which is also more than the combined market value of several head start-ups in the speech recognition industry (Spitzer, Yunzhisheng, go out and ask, Turing Robot, Jetong Huasheng) plus the market value of iFLYTEK, the voice leader of AI.
According to IDC,2017, the total size of China's computer application market is about 1.55 billion yuan. Shangtang, Kuangshi, Yitu and Yuncong have a total market share of 69.4%. By 2022, China's computer vision application market will reach 146yuan. 100 million yuan.
In the market pattern, technology giants control the basic layer, while start-ups take the lead in the application layer.Computer vision architecture from bottom to top:
1) basic layer-core chips are controlled by traditional chip manufacturers such as Intel and Nvidia, but new chip manufacturers have not yet emerged, and large-scale applications will take time; open source platforms are based on Alphabet Inc-CL C's Tensorflow,
Facebook Inc's Caffe is the main one, and most of the deep learning frameworks of other enterprises are secondary development.
2) the technology layer, algorithm, is dominated by startups; cloud computing is almost monopolized by AWS, Google Cloud, Azure, Aliyun, etc.
3) the application layer-the vertical industry leader occupies the scene, and the technology layer start-ups infiltrate upward.
At present, Chinese computer vision algorithm companies are ahead of their international counterparts in terms of technical level, degree of commercialization and financing ability (valuation).Technically, in recent years, Chinese computer vision head algorithm companies have frequently won the crown in international image recognition competitions such as ImageNet, FRVT, MSR IRC, etc. In commercialization, according to public reports, Shangtang, Xiangshi and Yuncong have all made profits in 2017, while their international start-up counterparts such as EverAI are still in the net investment stage; in addition, from the perspective of valuationThe four Little Dragons of CV have all reached a multibillion-dollar valuation, while there are no foreign start-ups with a valuation of more than one billion dollars based on face recognition and image recognition alone.
From the perspective of market structure, the domestic and foreign market structure is slightly different: domestic security, finance, Internet-based, foreign consumption, robot (and machine vision), intelligent driving.