floorballstudio
FI|EN
AI + digital coaching experiments

SHOT ANALYZER

Analyze floorball shooting mechanics from video.

Evaluate wrist, slap, drag, snap, backhand and one-timer mechanics by combining body pose, stick/blade detection and phase-aligned reference comparison.

Technology

MMPose or MediaPipe for body/hand landmarks; MMDetection + SAM 2 for stick/blade/ball; OpenCV for video; DTW-style alignment; optional Pose2Sim/OpenSim for 3D.

How a coach can use it

  • Identify contact frame and follow-through.
  • Measure knee, hip, trunk, shoulder, elbow and stick angles.
  • Compare timing and body position with multiple reference shots rather than one “perfect” player.
  • Turn findings into 2–3 concrete coaching cues.

Camera setup

  • Quick side view: phone 4–6 m from player, roughly hip/chest height, fixed, full body + stick + ball visible, preferably 60 fps or higher.
  • Front/rear 45°: useful for grip width, shoulder rotation and lateral movement.
  • Advanced two-camera mode: side + 45° view, both fixed and synchronized.

Implementation path

Start with a small browser prototype, keep the data model explicit, then connect the analysis backend as separate API services. Validate results with coaches before automating scoring.

Prototype

Try the prototype

This is a front-end concept demo. It does not yet send video or personal data to a server.