Action detection precision

[haildceu1]

Inspiring work! I‘m currently doing similar work except the scenario is on tennis. I'd like to know what action precision you guys can achieve by using YOLO-V8, considering that no temporal features are involved. Also, have you guys heard of SwingVision? What algorithm do you think this app uses cause it looks pretty amazing. Much appreciation!

[masouduut94]

Have a look at these results:

https://github.com/masouduut94/volleyball_analytics/wiki/Deep-Learning-Models#action-detection

[haildceu1]

Have a look at these results:

https://github.com/masouduut94/volleyball_analytics/wiki/Deep-Learning-Models#action-detection

Thank you. I'm also wondering what kind of GPU do you run the algorithm on. As far as I know, VideoMAE is quite large.

[masouduut94]

I trained the VideoMAE model with RTX 3090 24 GB. It doesn't need big GPUs

…

On Sat, Jul 27, 2024, 10:42 AM Zhuang Yiwen ***@***.***> wrote: Have a look at these results: https://github.com/masouduut94/volleyball_analytics/wiki/Deep-Learning-Models#action-detection Thank you. I'm also wondering what kind of GPU do you run the algorithm on. As far as I know, VideoMAE is quite large. — Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFB4FCDS3GHINQASZIP7V2DZONB5PAVCNFSM6AAAAABLOQCELKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENJTHA3DQOJRGU> . You are receiving this because you commented.Message ID: ***@***.***>