CICC RESEARCH: COMPUTER VISION APPLICATION LANDSCAPE IN CHINA

September 6, China International Capital Corporation Limited (CICC, 3908.HK), China’s oldest joint-venture investment bank, issued an analysis report on the landscape of computer vision industry applications in the China market. Huang and Yang, the authors of the report, stated that they believe computer vision applications are the most useful applications of AI in China, it has the most depth and breadth of user scenarios. Here is an excerpt from the report.

Investment Advice

Compared to language processing, computer vision applications experienced a better landing in the industries. From the current progress of computer vision technology implementation, the leader is in face recognition for security, and followed by retail and logistics tracking, while the medical imagery and self-driving cars are still a work-in-progress. The report noted, leading AI players, such as SenseTime, Face++, and Yitu Tech, are differentiating their strategies by either building enterprise platform or pursuing hardware-software integration. In addition, Clobotics and Deepblue Technology, the emerging players in this field, are also seeking smart hardware applications in new retail industries. The report advises investors to monitor the industry growth trends closely.

Figure 1. Landscape of Computer Vision Applications in China

Differentiation in Strategy

The report noted the differentiation in strategy-setting for the most significant players in the computer vision (CV) field in China. In the past year, it is starting to be more explicit that a) SenseTime is building the algorithm platform for enterprises, b) Face++ is going after various vertical industries implementation, whereas c) Yitu Tech is giving full play to their advantages in Cloud, focusing in the security and medical industries. It is yet too early to say which strategy is the most effective.

Figure 2. Data Accessibility and Technical Difficulty of CV Applications (The darker the color, the higher accuracy is required)
Figure 3. Clobotics’ Unique Positioning in the China CV Landscape

Predictions

In scenarios such as social media network, consulting, gaming, e-commerce, and security/access control, the error tolerance levels are relatively higher, and data are relatively easier to obtain, the computer vision applications were given room to multiply. On the other hand, self-driving cars and medical imagery recognition require very high accuracy, and the data complexity is exceptionally high, they believe in the short-term it is unlikely to achieve large-scale commercial use. The analysts expect the next wave of growth will be in retail, logistics, and manufacturing.