This technology open day is not simple. Let's take a look at the research results of keep in artificial intelligence. This technology open day is not simple. Let's take a look at the research results of keep in artificial intelligence. On December 22, the first technology open day of keep was held in Beijing, Keep technology VP pengyuehui and keep chief scientist qinzengchang gave a keynote speech on the evolution and development of keep technology, the application of AI technologies such as visual algorithms and big data mining in the sports field. The keep technology team demonstrated the representative actual combat cases of Israel on site, and discussed the landing and practice of AI cutting-edge technologies in the sports field with many technology lovers and technology masters in the artificial intelligence field, The on-site interaction atmosphere was warm, and the participants also actively shared their own experience in practice
keep technology evolution path to create an intelligent sports technology ecology
keep currently has a wealth of sports category services, providing 160million users with one-stop sports solutions such as fitness teaching, running, cycling, making friends, healthy diet guidance and equipment purchase. A powerful data center records users' sports behavior and data in an all-round way. In the process of operating the sports community, keep has also explored many KOLs in the sports field. It can be said that keep has become an important part of everyone's sports lifestyle
the exploration of science and technology has always supported the progress of keep. Pengyuehui, VP of keep technology, mentioned during the on-site sharing that the positioning of keep technology team is to continuously improve product quality and development efficiency, continuously strengthen the ability of technical support business, break through technical barriers and explore business value for keep. To this end, the keep technical team structure is also divided into three levels. The lowest level is the infrastructure and quality department, which provides support for all businesses and platforms above; The second layer is the middle stage, including commercial middle stage, tool middle stage, data and algorithm; The top level is commercial R & D, including tool R & D, community R & D and hardware R & D. The big data platform built by keep records the movement data generated by users, and further helps users improve the movement effect through "learning, practicing and testing". At present, keep has the most comprehensive and accurate data content in the whole movement field
in the aspect of AI, the current planning of keep mainly has two aspects: one is to explore the combination of motion and AI based on sensors or cameras, score the user's motion trajectory and give real-time feedback; Second, based on the massive data collected in the keep app application and intelligent hardware, it provides personalized sports guidance to users
after financing this year, keep raised AI to a strategic level and launched AI intelligent virtual coaching program. In order to promote the implementation of intelligent sports, pengyuehui introduced, "We have divided the whole process into four steps. The first step is to build scenes, such as the keepkit of family sports scenes and the keepland of urban sports scenes, even after years of exposure. The second step is to build content and give users a complete course system. The third step is to mine data and summarize the data through their own data warehouse, so that users' portraits and labels can be clearly known. The fourth step is to support algorithms, including For content distribution algorithms, hardware algorithms, visual algorithms and data mining algorithms, we hope that the data processing and application of the algorithms will finally feed back to users, so that users can get a better experience. "
at this stage, AI technology has been widely used in the sports field. Keep has continued to cultivate in the IOT and AI fields, and has made many attempts. It will not only provide users with personalized training content, but also help keep build a more perfect sports science and technology ecology
with user experience as the core, the core application of AI technology in the sports field
with the development of mobile Internet, we can enjoy the convenience of life without leaving home. Many people like to pay more attention to mental development than physical development, but in fact, health is very important to everyone. Maintaining physical health requires personal perseverance and the assistance of science and technology. Many people want to be active, but many methods are not scientific. People who really understand sports need to pass these information to the public. What is the most convenient and efficient way? Or through information technology, mobile Internet and artificial intelligence
keep hopes to solve some direct and key problems for users. Customized training is an important thing keep is doing. Qinzengchang, chief scientist of keep, told the technical practitioners on the spot that, "We have collected a wealth of user data. Through data intelligence, we can provide customized intelligent training plans for users. We will turn the professional knowledge given to us by all coaches and the movement goals achieved by these users into a probability model. Through the analysis and calculation of determining the limiting conditions, we will sample a path in the entire solution space. This path is a 'personalized' training scheme for this user."
everyone's needs and physical conditions will constantly change during exercise. By connecting with keep's intelligent hardware, you can timely adjust the user's training program. For example, when users wear bracelets and exercise, they feed back information to us through some continuous signals. Keep will calculate, analyze and recognize these signals as user actions. The technical exploration in IOT is a major sports technology field that the keep AI team focuses on. This year, keep released several hardware products, such as treadmill and body fat. Keep connects the data of intelligent hardware with the user's ID. after establishing a data connection channel, it can more conveniently provide users with more timely motion feedback. In the later stage, keep will also show the users that the experimental machine can not operate because of the damage of a local part in many times during movement through the function of TOF camera, so that users can understand their own physical conditions. Finally, qinzengchang further shared the vision exploration of keep, "At this stage, keep is not only working on popular content detection algorithms, but also focusing on the recognition of key actions and gestures, as well as the detection of key points of the human body. For the form of pictures and texts, we can do some cross media search work. For example, we can put forward a prediction for pictures and words, turn the original text into an expressive vector, and In terms of the comprehensive information of the text, the graph comprehensively represents how to measure the relationship between semantics. Basically, it uses the information from the picture and the embedded information from the text to establish a model to realize this part of the research work. "
keep AI research institute's current research mainly has four directions: one is machine vision, which can get more feedback on people's body and actions; the other is data mining content distribution, including modeling to provide more distribution and semantic association; the third is the construction of big data bottom layer and data mining; the fourth direction is intelligent hardware algorithm, and keep will also pay attention to the research progress of other AI
speak with technical strength. Demo applications show "future sports life"
on the open day, the keep technical team showed us a lot of cool demo applications on the spot, such as using visual algorithms and CNN to detect the body posture of each user, and taking photos of the movement process through the TOF camera to get action guidance and real-time feedback. These things seem to be accomplished by algorithms, Keep will take them to the extreme and apply them to apps in the future
in fact, many applications in the demo need to be deployed to the mobile terminal. Therefore, keep shows a fast and lightweight human body recognition method. For example, the function of the bracelet to monitor rope skipping is mainly based on the six axis or nine axis sensors in the equipment to collect data and form judgment standards according to the motion features extracted during rope skipping. In addition, keep also uses the constructed algorithm system to detect the key points of the human body, so as to monitor the action standard of the user in the process of movement, and generate the motion style feedback to the user
it can be seen that improving the user experience and providing practical services for users are the key to the application of technology, and keep's exploration in technology is constantly evolving to meet the needs of users
in the form of a technology open day, keep hopes to integrate and launch valuable technology resources, enable more technology practitioners to share and apply them, and link the technology industry. In the future, Kee with poor anti-interference capability will continue to launch more forms of industry technology sharing and exchange activities. Let's look forward to keep bringing more surprises