● LIVE
sensei.coach
Video Analysis

Joint kinematics. Movement quality. No fabricated counts.

33 skeletal landmarks per frame. 14 joint angles, 3 axial rotations (pelvis, shoulder girdle, X-factor). When confidence is low, Sensei refuses to invent strike counts — and tells you so.

01
The pipeline
Self-contained · runs on Sensei infrastructure

Your video never leaves our pipeline.

We use Google's MediaPipe BlazePose Heavy model (pinned to a specific build, float16, downloaded once at container build) running in VIDEO mode for temporal consistency. No DeepMotion, no Move.ai, no third-party motion-capture API.

33
skeletal landmarks
14
joint angles
3
axial rotations
EU
hosted · Ireland
02
Confidence score
Four factors · surfaced to the user

Every analysis ships with its own confidence pill.

FactorWeightWhat it measures
Temporal coverage35%Fraction of usable frames out of frames sampled.
Mean frame confidence30%BlazePose per-landmark certainty, averaged.
Stability20%Fraction of usable frames that are consecutive.
Resolution15%Source resolution as a 0–1 score.
HIGH
≥ 75

Full frame-anchored coaching with strike-level reads.

MEDIUM
50 – 74

Specifics with hedged language and visible confidence pill.

LOW
< 50

Sensei refuses strike counts; offers qualitative observations only.

03
What we don't claim
Hard constraints we're proud of

Honest about what the camera can't see.

  • We don't claim accurate strike counts. We talk about joint kinematics and movement quality instead.
  • We don't promise a specific time-to-feedback. A 30-second clip typically takes 45–90 seconds end-to-end on EU infrastructure.
  • We don't do live mobile-camera analysis. Upload-only today; real-time is on the roadmap and will be flagged when it ships.
  • We don't make medical claims. If something hurts, see a human.
● Try it

Drop a clip. See it honestly.

Two video uploads a month, free.

Upload your first clip →Back to home