The Volleyball Scientist

volleyball 🏐📊

[The Volleyball Scientist @LearningVolley]

Here's a D1 NCAA Women's Volleyball ranking based on points added/saved per set.

Offensive rating = points scored above average.
Defensive rating = points opponent scored below average.

Ranking them with offensive + defensive as a "net impact" score.

[The Volleyball Scientist @LearningVolley]

Mayu Ishikawa goes over top Anna Smrek's block, who is 30 cm (1 ft) taller

[The Volleyball Scientist @LearningVolley]

no setting point for 11. yamada on that? :/

[The Volleyball Scientist @LearningVolley]

Japan's middle blocker digging the net ball to the outside

[Gibby @gibbybaris]

⚡️ Yoshino Sato x Ukrayna ( 4 Set )

17/46 %37h, 3 hata, 3 bloklanma
2 blok, 3 servis hatası
%68 pm, %36 mm ( 25 manşet )
19 sayı, +10 WL

[The Volleyball Scientist @LearningVolley]

dominican republic won the first set.

[The Volleyball Scientist @LearningVolley]

brazil came back

[The Volleyball Scientist @LearningVolley]

brazil came back

[The Volleyball Scientist @LearningVolley]

ukraine beats germany
netherlands beats turkey
and now the dominican republic?

[The Volleyball Scientist @LearningVolley]

ukraine beats germany
netherlands beats turkey
and now the dominican republic?

[The Volleyball Scientist @LearningVolley]

@hexwanderer @tejfbanalytics tale as old as time

[Winston @hexwanderer]

@tejfbanalytics Where are people getting these datasets. UI is easy

[The Volleyball Scientist @LearningVolley]

@markpinc That's cool how do I get my Stem Engine game featured in the docs :)

[The Volleyball Scientist @LearningVolley]

@roboflow coudln't have come at a better time

[The Volleyball Scientist @LearningVolley]

@MLVMinnesota Nice!

[The Volleyball Scientist @LearningVolley]

@MLVMinnesota How did they find the exact photo for Samedy

[The Volleyball Scientist @LearningVolley]

Cody Miller looks huge at the Enhanced Games

[The Volleyball Scientist @LearningVolley]

here's the video with noisy players.

[The Volleyball Scientist @LearningVolley]

finding volleyball players

[The Volleyball Scientist @LearningVolley]

btw YOLO is doing about 30 fps for me. It does a good job on close players but I can't reliably track the opponent.

Tomorrow I'll add ball tracking and we should be close to something like setter angle.

This video is a little fisheye lens though - the baseline is curved

[The Volleyball Scientist @LearningVolley]

finding volleyball players

[The Volleyball Scientist @LearningVolley]

This is with SAM3 which processes at about 1.5 frames per second on my machine.

You could do this live with YOLO but it'd have to be in a less noisy/practice setting.

YOLOv8:

[The Volleyball Scientist @LearningVolley]

After a few days of image annotation I think these models like SAM3 are already so strong that fine-tuning the models isn't as important as the post processing...

[The Volleyball Scientist @LearningVolley]

@Meta Well the amazing part it's not including the crowd. That's without any priors on player uniforms - I mean the people in the stands are wearing the same color as the players!

And you don't care about bench player's labels. you only care about correctly labeling active players.

[The Volleyball Scientist @LearningVolley]

4/ I'm back from vacation and I'll be spending the next day labeling! I wanted to share how strong image classifiers are right now.

Using SAM3 (@Meta's open source Segment Anything Model) I'm able to very quickly label these images. Maybe I shouldn't even be labeling these...