36 Kr starter | "Galaxy Waterdrop" received 300 million yuan of B round financing, and gait recognition entered the stage of rapid commercialization

Author | Chant

Edit | Su Jianxun

36Kr learned that Galaxy Waterdrop Technology (Beijing) Co., Ltd. (hereinafter referred to as "Galaxy Waterdrop") has completed nearly 300 million yuan of B round financing, and this round of financing is led by Haitong Securities and Jiuzhao Investment, and Zhongke Fusion Innovation Accelerator serves as the financial advisor.

New technologies such as AI are often used in various identity authentication scenarios, the most common of which is face recognition, fingerprints, irises and so on, which are already common. These technologies have been used for a long time and have the uniqueness of recognition, but the limitation is that they can only be recognized at close range.

In contrast, gait recognition, the new AI technology direction in recent years, refers to the analysis of people by body size and walking posture, and its physical basis is the different body physiological structure and walking habits of each person. For example, the face or iris have requirements for lighting, face recognition will be seriously affected at night, but the gait is not sensitive to light, and the identification object is difficult to camouflage, no sense when identifying, no need to align the camera, which can become a supplement to long-distance identity authentication.

Galaxy Waterdrop is an AI company that focuses on gait recognition technology. After four generations of doctors and 20 years of theoretical and application experience, its founding team is the first company in China to devote itself to commercial gait recognition technology.

Identity authentication technologies such as face and fingerprint were first landed, and their training methods and objects are closely related. First of all, these features are mainly based on image training, and the rapid development of terminal equipment such as mobile phones makes the acquisition of images/fingerprints extremely common, and the sample size is huge. Compared with this, gait recognition can only be recognized by video, and it is affected by weather, character dress, character angle, etc., and the early recognition accuracy is low, and commercialization has been limited.

Since its inception in 2016, Galaxy Waterdrop has been focusing on the research and development of gait recognition technology, and so far, the company has entered a new stage - the technology is becoming more and more mature, and after this round of financing for rapid commercialization. "After a long period of investment, the current gait recognition technology of galaxy water droplets has fully met the needs of actual combat landing." Founder and CEO Dr. Yongzhen Huang told 36Kr.

Whether an AI direction is mature, the training base library is the core indicator, and the mature commercial technology is at least based on the million-level training library. In the past two years, Galaxy Waterdrops has made rapid progress at the algorithmic and technical levels. In terms of recognition accuracy, the Galaxy Water Drop is currently undergoing gait retrieval under the million-level public security combat bottom reservoir, and the first life rate has far exceeded 90%.

At present, Galaxy Waterdrop has taken the lead in proposing a new video infrastructure strategy with gait recognition as the core to solve the pain points of China's massive video analysis, and its core product "Vision" system has three major sections: full element analysis, full scene recognition and diversified services.

Huang Yongzhen introduced to 36Kr that from the perspective of product positioning, "full element analysis" and "full scene recognition" refer to covering 90% of the cameras and parsing 90% of the targets, and the target coverage of the video is huge. Taking the security scene as an example, due to too many surveillance screens, public security personnel can not stare at the screen at all times, most of the videos are invalid "zombie videos", galaxy water droplets are hoping to eliminate such "zombie videos", solve the problem of pedestrian identification in complex scenes, so that videos can really be effectively applied to various scenarios.

In the process of technology research and development, Galaxy Waterdrop has also broken through a number of difficulties in gait recognition and established high barriers in the field of video-based identity authentication. At the same time, Galaxy Waterdrop is also the company that led the development of China's first gait national standard.

"For example, the blurred vision of the camera when it is windy and rainy, and the change of people's clothing, are the difficulties of gait recognition, and compared to flat faces, gait recognition is a full range of 360 degree recognition, and recognition at different angles is a difficult point that has plagued the entire gait recognition field for decades, and these galaxy water droplets are breaking through one by one." Huang Yongzhen said.

In terms of commercialization, Galaxy Waterdrop has also established a number of benchmark customers in the fields of public safety, auxiliary medical care, traffic analysis, educational examinations, and sports training. This also further trains the landing ability of galaxy water droplets, and at present, galaxy water droplets have been able to reach the concurrency of thousands of high-definition cameras, and will further cooperate with chip companies to improve computing power.

Public safety is the key landing scene of the Galaxy water droplet. Taking a second-tier city as an example, after deploying the thousands of vision systems of Galaxy Waterdrops, the vision system was launched for one month, assisting the public security to crack more than 20 cases, and the number of cases of robbery and pickpocketing dropped by 66% and the number of cases filed fell by 25.9%. In the medical field, gait analysis can give auxiliary diagnostic suggestions for patients' physical abnormalities and rehabilitation by analyzing stride length, stride speed, and joint movement angle, and can be applied to orthopedics, neurosurgery, rehabilitation, etc.

At the platform level, the current Galaxy Waterdrop has expanded to multiple industries and horizontally expanded to multiple modules, including AI middle office, technology middle platform, etc. In each subdivision of the industry, Galaxy Waterdrop will also take into account the needs of different segments, such as the field of public safety, in the past two years, Galaxy Waterdrop has subdivided into criminal investigation, anti-fraud, stability maintenance, major project security and other directions of solutions. At the revenue level, Huang Yongzhen said that Galaxy Waterdrop is currently dominated by large customers, and it is expected that there will be a big breakthrough in revenue this year.

In the future, Galaxy Water Drops will focus on the larger level of "video new infrastructure" field.

Previously, China invested trillions of dollars to build a "video old infrastructure" - the deployment of massive cameras and networking platforms, etc., in an attempt to solve the problem of digitalization of physical space, thereby improving the efficiency of social governance and services, typical scenarios such as chase and escape, counter-terrorism, child abduction and retrieval, the elderly lost and recovered, epidemic prevention and circulation, etc.

"The original infrastructure is mainly to solve the problem of terminal equipment deployment and networking such as cameras, such as the Skynet Project has deployed more than 20 million high-definition cameras." In the past two years, the country has begun to mention the concept of ai new infrastructure, which is a broad space for development for Galaxy Waterdrop. Huang Yongzhen said.

However, the old video infrastructure has several pain points, basically more than 90% of the cameras have become zombie cameras, because the number of videos is too large, it is difficult to process with the naked eye; in addition, more than 90% of the targets cannot be analyzed, because the outdoor scene and environment are very complex, and the angle, distance, dress, lighting and other issues bring great challenges to the AI algorithm.

Huang Yongzhen believes that the solution of Galaxy Waterdrop with gait recognition as the core technology will greatly promote the efficiency and intelligent level of social governance and services on the basis of old infrastructure in the future. The focus here is on moving from digitalization to the intelligent stage - such as how to analyze and connect the existing massive videos, how to further radiate to more partners through the scene, so as to tap the new needs of customers, which is the focus of the development of Galaxy Hydropower.

At present, Galaxy Waterdrop has a team of more than 200 people and is headquartered in Beijing. About 50% of the technical team is a doctor of the Chinese Academy of Sciences, and more than 80% has graduated from scientific research institutes and universities such as the Chinese Academy of Sciences, Tsinghua University, and Peking University.

Galaxy Waterdrop announced the completion of a 100 million yuan pre-a round of financing, led by Hengbang Capital and Huahe Capital

Turning the sci-fi scene of "Mission Impossible 5" into reality, "Galaxy Water Droplet" uses gait recognition to achieve identity authentication