# What is the best pose detection and motion tracking?

**URL:** <https://community.ultralytics.com/t/what-is-the-best-pose-detection-and-motion-tracking/1982>\
**Category:** Discussion\
**Tags:** discussion, code, ai, yolo\
**Created:** [May 21, 2026, 10:17am UTC](https://community.ultralytics.com/t/what-is-the-best-pose-detection-and-motion-tracking/1982 "2026-05-21T10:17:46Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Curtis](https://avatars.discourse-cdn.com/v4/letter/c/58956e/32.png) [@Curtis](https://community.ultralytics.com/u/Curtis)\
**Post date:** [May 21, 2026, 10:17am UTC](https://community.ultralytics.com/t/what-is-the-best-pose-detection-and-motion-tracking/1982/1 "2026-05-21T10:17:46Z")

</div>

I am 100 videos deep in individually pausing and pressing play every few seconds of a 18 second video of a surfer surfing in a wave pool. I have created a formula for a scoring output for each individual manoeuvre and action during the clip of the surfer on the wave. I am using the 33 key point skeleton extraction over the body for manoeuvre position tagging. What is the best product out there for ai vision tracking and pose detection ? My detection is slow and lagging, it’s okay but not great

---

<div class="post-metadata">

**Author:** ![pderrenger](https://sea1.discourse-cdn.com/flex001/user_avatar/community.ultralytics.com/pderrenger/32/73_2.png) [@pderrenger](https://community.ultralytics.com/u/pderrenger)\
**Post date:** [May 22, 2026, 1:19am UTC](https://community.ultralytics.com/t/what-is-the-best-pose-detection-and-motion-tracking/1982/2 "2026-05-22T01:19:00Z")

</div>

If you want the best overall balance of speed, accuracy, and deployment right now, I’d use [Ultralytics YOLO26 pose](https://www.ultralytics.com/blog/ultralytics-yolo26-vs-other-ultralytics-yolo-models-for-pose-estimation). It’s the current recommended Ultralytics model for pose work and is a strong fit for real-time sports analysis.

One important detail: if your pipeline depends on a 33-point body skeleton, that usually points to MediaPipe-style landmarks. The standard pretrained YOLO pose models are typically 17-keypoint human pose, so for your surfing scoring workflow the best path is often a custom-trained YOLO26-pose model on your own 33-keypoint labels. That usually matters more than the model name alone, especially with wave-pool angles, spray, occlusion, and fast turns.

If you want, I can help you pick the fastest setup for your videos, either:  
`MediaPipe for fixed 33 landmarks`  
or  
`YOLO26-pose custom training for better tracking/robustness`