Group Actions
Engagement
  • Author Author: albertabeef
  • Date Created: Date Created
  • Views 3610 views
  • Likes 7 likes
  • Comments 11 comments
Related
Recommended

Applying AI to climbing … The Long Road to the 2020 Tokyo Olympics

albertabeef

In this blog series, I attempt to bring two of my passions together … AI and rock climbing.

image

 

2020 was the year of Rock Climbing, the first Olympic event to feature rock climbing as a discipline.

Unfortunately, circumstances have prevented that from happening.

I was pleased to hear that the 2020 Tokyo Olympics will finally take place from July 23 to August 8, 2021, although no overseas spectators will be allowed due to covid-19 risks.

If you are new to rock climbing, you definitely have to check it out.  The moves that some of these athletes accomplish are absolutely incredible !

 

The broadcast industry is always innovating to improve the user's immersion into these fascinating sports.

What can we expect for climbing ?

Will they help the audience understand the subtleties of this new sport, and share with us the signature moves such as "dyno", "gaston", "heel hook", "knee bar", "toe hook" ?

 

Rock Climbing MovesExample Images
Back Flagimage
Bat Hangimage
Dynoimage
Gastonimage
Heel Hookimage
Knee Barimage
Lockoffimage
Mantleimage
Stemmingimage
Toe Hookimage

 

Let's not forget some of the moves from the related sport of ice climbing

Ice Climbing MovesExample Images
Leg Rest Over Armimage
Leg Rest Over Opposite Armimage

 

 

 

Will they be able to show us side-by-side comparisons of different climbers ?

image

 

Often times, different climbers will find different solutions to the problems.

 

The routes that we climb are called “problems”, for which we ultimately need to find “solutions”.

For very advanced climbers, the ability to resolve a “problem” is part of the experience, and they prefer to figure it out themselves.  This is because they have extensive technical abilities, and know how and when to use specific moves.

For the novice or average climber, figuring it out can sometimes turn out to be insurmountable.  Enter the “beta”.  “Beta” is information regarding the use of a specific move on a particular section of a route (or problem).  Climbers generously share their “beta”, when asked.  Since all climbers are different, sometimes one climber’s “beta” will not work for another climber, due to height, skill, etc…

In the Olympics, there is no access to "beta", all climbers must figure out the "problems" based on their experience, technique, and ingenuity.

 

There will be 20 women and 20 men competing in the 2020 Tokyo Olympic games.

Here is a list of the athletes that have qualifies thus far (in alphabetical order of last name)

  • Women
    • Julia Chanourdie (France)
    • Kyra Condie (United States)
    • Shauna Coxsey (Great Britain)
    • Janja Garnbret (Slovenia)
    • Iuliia Kaplina (Russia)
    • Petra Klingler (Switzerland)
    • Mia Krampl (Slovenia)
    • Aleksandra Miroslaw (Poland)
    • Akiyo Noguchi (Japan)
    • Miho Nonaka (Japan)
    • Jessica Pilz (Austria)
    • Brooke Raboutou (United States)
    • Laura Rogora (Italy)
    • YiLing Song (China)
    • Alannah Yip (Canada)
  • Men
    • Nathaniel Coleman (United States)
    • Colin Duffy (United States)
    • Ludovico Fossali (Italy)
    • Kai Harada (Japan)
    • Jan Hojer (Germany)
    • Rishat Khaibullin (Kazakhstan)
    • Alberto Ginés López (Spain)
    • Bassa Mawem (France)
    • Mickael Mawem (France)
    • Sean McColl (Canada)
    • Alex Megos (Germany)
    • Tomoa Narasaki (Japan)
    • Adam Ondra (Czech Republic)
    • YuFei Pan (China)
    • Jakob Schubert (Austria)

 

Share you feedback.

  • What country are you from ?  Will your country have an athlete competing in this event ?
  • Do you rock climb ?  Do you ice climb ?
  • Who is your favorite climbing athlete ?
  • What are the three disciplines that are part of the climbing event in the 2020 Tokyo Olympics ?

 

In my next blog, I will investigate how standard deep learning models work with climbing images and videos.

 

 

 

References:

Parents

  • Hi Mario,

     

    I'll be interested to see how well you do with this.  My first thought is that you'll need a lot of image data to be able to discriminate the various moves.  Is there an image database for rock climbing?

     

    Maybe initially you could identify the route and based on other parameters (time, climber characteristics), you could identify which videos are of interest (and maybe which points are of interest) - and then try to focus on identifying a particular move at a particular section.

     

    Almost seems like too hard a problem to solve with a 2D video image.  It will be fun to see your "beta" for this problem image.

     

    -Ralph

Comment

  • Hi Mario,

     

    I'll be interested to see how well you do with this.  My first thought is that you'll need a lot of image data to be able to discriminate the various moves.  Is there an image database for rock climbing?

     

    Maybe initially you could identify the route and based on other parameters (time, climber characteristics), you could identify which videos are of interest (and maybe which points are of interest) - and then try to focus on identifying a particular move at a particular section.

     

    Almost seems like too hard a problem to solve with a 2D video image.  It will be fun to see your "beta" for this problem image.

     

    -Ralph

Children