Climbing tool uses VLM agent for hold detection and pose estimation visualization

jonstephens85 · x · 2026-08-26

A developer created a rock climbing tool using @vlmrun's visual agent Orion. By prompting the VLM, the system segments blue bouldering holds without training a custom dataset. Combined with pose estimation, it visualizes how each hold is activated by hands and feet, tracking the torso's path to help climbers analyze their movement.

Original post →

More from Multimodal

Multimodal channel →