Pose Detection

Problem

Fitness instructors, choreographers, and movement analysts want to compare poses across video clips without frame-by-frame manual inspection.

Solution

FlexGrid uses Vision framework body pose detection to extract 19-joint skeletal fingerprints from video frames. Users can save poses to a Saved Pose Library, then use the Pose Compare Loupe to overlay and compare poses side by side. Similarity is scored using cosine similarity between joint vectors.

UI Flow

1. Right-click a video cell and choose “Detect Pose.”
2. FlexGrid overlays the detected 19-joint skeleton on the current frame.
3. Save the pose to the Saved Pose Library with a name.
4. Open the Pose Compare Loupe to select two poses for comparison.
5. The similarity score (0-100%) displays alongside the overlaid skeletons.

Tier Availability

Pro only.

Known Limitations

• Requires a clearly visible single person in frame; multi-person detection is not supported.
• Accuracy decreases with low-resolution or heavily occluded subjects.

Marketing One-Liner

Compare body poses across clips with skeleton detection and cosine similarity scoring.