Low and not-so-low hanging fruit for ovml

[ick003]

Hello everyone, I meant to post this a short while ago and got sidetracked. I would like to list any ideas as to what we could use AI-vision for in terms of volleyball analytics, and/or more. I believe that if we can showcase something really useful, and explain it through a blogpost (for example) we might have a better outreach. AI-vision is kind-of an ideal setup because pictures are more attractive than statistics tables (well, I suppose so)... ;)
So I'll go first with my ideas, I'd be happy to see you guys comment on them, willing to start working on them, or provide other ideas to explore!

• 1: Analyse the relationship between the setter position at the time of ball contact and where the set goes (Front, Back, Center, Pipe). We could use an appropriate ML alg (for example Keypoints R-CNN, see image attached) and a 3D or 4D Wasserstein Barycenter to identify whether there is a position pattern. It would require manual tagging to start with, and selecting a specific setter to make an example but I think it would be real nice, and useful!
• 2: Along the same lines, identify hard swing vs tip for a hitter, or any other variations of technical skill segmentation
• 3: If some of you are familiar with the Wii (although I suspect all videogame platform offer some of it) a funny application of Keypoints R-CNN would be to define a model for a given skill (say arm swing) and propose to compare live during a training session a set of "ideal" swings to what is actually perform by the (beginner) player.

There you go!

[tyler-widdison]

Nice ideas Adrien! I'll give it a simple whirl for now...
For beach:

Breaking down a teams defense plays has been something I've been playing with. It is possible to a certain extent within DVWs. Looking at 'fakes' (fake line and block angle) and the movement between the blocker and defender at a given point.

Each team will defend their opponent(s) differently. But some teams will have 'go-to' defense patterns they use. The patterns a team will use vs one specific team could provide insight where a team needs to improve on offense.