From: IBM By: Ace
Twelve years ago, the American writer David Foster Wallace wrote the article " Federer , a religious experience" in the New York Times. He began with this: "Looking back at the past. Over the past few years, almost every tennis fan and person who follows the men's tennis scene has experienced some classic moments that we call 'Federer Moments.'" Such moments were truly played out at this year's Wimbledon arena. , the nearly five-hour battle of the century between Federer and Djokovic made countless tennis fans dance with joy or sadness.
Although Federer missed his ninth Wimbledon championship and his 21st Grand Slam after a long set and three tiebreaks, the 38-year-old is still a classic legend in the tennis world.
IBM has provided a lot of technical support as one of the sponsors of Wimbledon in the past 30 years. As early as the era when Internet access still required dial-up, it helped develop and launch the first version of Wimbledon's official website. Now it uses its artificial intelligence platform Watson to help Wimbledon events automatically capture exciting video excerpts. has greatly improved the fans' viewing experience and The media's reporting efficiency and production of on-field analysis videos for each player within 20 minutes help players discover and solve problems as early as possible.
AI selected highlights on the court
Perhaps only loyal tennis fans will lie down in front of the TV to watch the five-hour tennis match in its entirety, and viewers who have been to the scene will know that most matches will be better than The excerpts on TV and the Internet are even more boring, and the highlights of the entire game may only add up to a few minutes.
Therefore, watching ATP and WTA official releases or fan-made HL videos is a more efficient way for tennis fans to watch the game. However, 13 days of competition, hundreds of tennis players, and 18 tennis courts are held at the same time for an average of at least 2 hours. It is very costly in terms of human resources and time to extract the essence of each game and publish it on the official website in a timely manner on the same day. High requirements.
IBM’s Watson AI platform can show its talents in this regard. Last year, Wimbledon generated a total of 2,000 edited videos, of which IBM AI contributed 14000,000. At IBM Wimbledon Tech Bunker, 180 on-the-job staff and AI systems expertly carry out all this.
Since 1990 years ago, IBM has collected 63 million data points on prize money from major tennis tours. After training on thousands of game videos, the system has become more and more proficient in selecting and ranking the best results in each game in real time. The exciting part. On what basis? Data points, including the real-time score on the field, the excitement of the players, and the actions taken after winning the score, AI needs to understand and identify these important moments, from the reaction of the audience to the posture of the players, are all the focus of the system direction.
Within 2 minutes after the game ended, before the dust had even settled, the system automatically and quickly edited the small clips together and released them to the entire network.
So how does the system learn excitement? First, IBM's AI needs to have a sensitive ear and hear some iconic sounds to decide where to start cutting, such as the cheers of the audience. Beginning with this year's Wimbledon, Watson Acoustics began to "listen" to the sound of the racket hitting the ball (the sound of a successful hit is crisp and bouncy) to increase the richness of the data. The sound analysis can now also identify the time when the ball is hit. , thereby editing out a highlight collection that combines sound and images more closely.
Secondly, it also needs a pair of keen eyes, using the Watson image recognition system to observe every movement of the players rather than the subtle expressions, because as you know, expressions are very different from person to person.
Visually recognized actions include clenching a fist when winning, breaking the racket when angry, shaking hands with an opponent, and many other gestures that represent specific meanings or occur in specific scenarios. But anything can happen on the court. The movement trajectories of a player walking towards the caddy to ask for a towel and waving his arms to win a ball and just walking towards the caddy may be similar in the eyes of the machine. Therefore, unless the system is trained in a very efficient way, it may assume that these two seemingly similar actions are the same thing.
However, live matches on different courts may bias the machine’s judgment. For example, compared with the women’s doubles match held on the Center Court, the former is better in terms of court quality and number of spectators. is better than the former, but this does not mean that the highlights of the latter must be less than that of the former.
In addition, the reaction of the live audience is also regarded as one of the indicators of whether the game is exciting, but if the audience's preferences are fully accepted without considering other factors, it may cause errors in machine judgment. In addition to fan preferences for specific players, unconscious patriotism can also influence the direction of normal games. Unless there are big stars like Sharapovawho are warmly welcomed by fans everywhere, generally speaking, domestic audiences are more inclined to cheer for domestic players. To solve this problem, IBM's OpenScale takes more factors into consideration, increasing the fairness of the results and reducing errors in highlights videos.
Of course, IBM Watson’s practice at Wimbledon also reflects some common problems in the implementation of today’s artificial intelligence technology. Although Watson can pick out the best candidate clips for highlight videos, to be sure, most of the key data - about 450 data points per tour - is still obtained by trained staff sitting on the sidelines using a The special tablet quickly records the results of serves, draws and points. Although some clubs and tours use computer vision algorithms to automatically generate data, sometimes with a 95% accuracy rate, IBM's executive responsible for the Wimbledon event said that a 5% error is not feasible for the four Grand Slam live events. tolerated.
As artificial intelligence plays an increasingly important role, sports venues have also become a battlefield for many technology giants. In addition to Wimbledon, IBM also maintains a similar cooperative support relationship with the US Open. Amazon AWS provides technical support for the NFL player tracking technology (Next Gen Stats). It is also the cloud service provider of NASCAR, Microsoft maintains a partnership with Real Madrid. Wimbledon 's ambition is to become the most prestigious tennis tournament in the world. We believe that the long-term cooperation with IBM will inject more new technological impetus into achieving this goal.
Refs:
https://www.digitaltrends.com/cool-tech/ibm-wimbledon-watching-artificial-intelligence-revolutionize-sports/
https://www.ibm.com/marketing/uk-en/wimbledon/
https: //www.nytimes.com/2006/08/20/sports/playmagazine/20federer.html
https://www.ibm.com/blogs/think/uk-en/ibm-serving-up-ace-ai-at- wimbledon-2019/
https://fortune.com/2019/07/11/wimbledon-2019-ibm-tech/
https://www.forbes.com/sites/stevemccaskill/2019/07/08/wimbledon-2019- how-ibm-addresses-crowd-bias-in-ai-generated-video-highlights/#4ab5384514c4
-The End-
Jiangmen is a new startup focusing on discovery, acceleration and investment in technology-driven startups The investment institution covers Jiangmen Innovation Services, Jiangmen Technology Community and Jiangmen Venture Capital Fund. Jiangmen was founded at the end of 2015. The founding team is composed of the original founding team of Microsoft Ventures in China. It has selected and in-depth incubated 126 innovative technology startups for Microsoft.
Jiangmen Innovation Service focuses on applying innovative technologies to real application scenarios, activating and realizing new business values, and serving industry-leading enterprises and technologically innovative startups.
Jiangmen Technology Community focuses on helping technologically innovative startups provide technology sharing and learning content from core technical experts in the fields of production, academia, research and innovation, making innovation a sustainable core competitiveness.
Jiangmen Venture Capital Fund focuses on investing in start-ups that activate business scenarios and realize commercial value through technological innovation. focuses on technology fields including machine intelligence, Internet of Things, natural human-computer interaction, and enterprise computing.In the past three years, Jiangmen Venture Capital Fund has invested in dozens of technology-based startups with high growth potential, including Quantitative, Malong Technology, Hesai Technology, Quantuo Technology, Shanshu Technology, Diyinga Technology, etc. company.
If you are a start-up in the technology field and not only want to obtain investment, but also want to obtain a series of continuous and valuable post-investment services, you are welcome to send or recommend projects to my "door": bp@thejiangmen.com
Click the upper right corner, Put the article in your circle of friends
Jiangmen Ventures
to let innovation be recognized!
WeChat : thejiangmen
bp@thejiangmen.com
