Analyze exercise form using in-browser pose landmarks and joint-angle computations.
npx skills add https://github.com/AmjadShakhshir/gaurdform --skill pose-form-analysis
Please help me install this Agent Skill. Skill: pose-form-analysis Source: https://github.com/AmjadShakhshir/gaurdform/tree/main Command: npx skills add https://github.com/AmjadShakhshir/gaurdform --skill pose-form-analysis
This Skill enables real-time exercise form analysis in the browser by leveraging MediaPipe Pose Landmarking to compute joint angles, smoothing, and a rule-based evaluation for common exercises like squats and bicep curls.
Run the Pose Landmarker setup, feed landmarks into the rule engine, and render live feedback in real time.
High-intent search queries and answers about installing and using this skill.
Real-time exercise form feedback in a PWA is achieved by running MediaPipe Tasks Vision PoseLandmarker in-browser, computing joint angles from landmarks, and applying a rule engine for smoothing and evaluation.
Squat and bicep curl rep counting uses state machines driven by computed joint angles from MediaPipe pose landmarks, tracking movement phases to increment repetitions accurately during live analysis.
Yes, an optional Gemini 2.0 Flash coach review layer can be integrated alongside browser-based pose detection, adding multimodal feedback including visual chips, speech, and haptic cues to the rule-based evaluations.
MediaPipe Tasks Vision supports deterministic front-end integration for on-device exercise form analysis, running pose landmarker computations directly in the browser for fitness PWAs without requiring backend processing.
The best way to smooth noisy joint angle data during live pose tracking is to apply mathematical smoothing algorithms to raw MediaPipe pose landmarks before evaluating them in the exercise rule engine.
damionrashfordRun MediaPipe Vision, Audio, Text, and GenAI tasks via CLI with structured JSON output.
marmikcfcGenerate images from text prompts via the Nano Banana API on Orthogonal.
TaylorHustonGenerate spaced repetition flashcards from Obsidian notes in multiple formats.
NVIDIA-TAOAutomate training, evaluation, and deployment of 6-DoF pose estimation models.
