HUMAN KINETICS
REDEFINED.
Personalised AI coaching built on MediaPipe pose streams, a PyTorch SAC policy for twelve joint corrections, a 22-class workout DNN (98% F1), and a seven-group muscle Random Forest (90% F1)—all with sub-millisecond model inference.
For Athletes
Move with data-backed precision. TrueForm reads your pose sequence, classifies the lift, scores muscle-group activation, and nudges twelve joint targets in real time so quality reps stack up and risky patterns surface early.
Track Progress
Real-time velocity and range-of-motion metrics for every session.
Optimize Training
Sequential models and ensembles prioritise corrections that match the workout you are actually performing.
Prevent Injury
Pose-based cues target long-term injury reduction; we are extending the same stack toward rehab and physiotherapy workflows.
For Coaches
The command centre for coaches who want the same signals the models see: MediaPipe overlays, class probabilities, muscle-group votes, and SAC-guided joint deltas—fed through a modular ML release pipeline.
-
sensors Remote Monitoring
Real-time sync of athlete sessions from anywhere in the world.
-
analytics Data-Backed Feedback
Overlay audio and visual cues directly onto athlete skeletal playback.
-
groups Client Management
Centralized dashboard for automated scheduling and progress reporting.
-
psychology AI Insights
DNN + Random Forest outputs plus SAC corrections surfaced as explainable cues on every rep.
Client: Sarah Jenkins
FORM DEVIATION
Protocol Intelligence
Latency budget
Regression and classification heads stay under a millisecond so coaches and athletes get feedback in the same rep, not the next set.
Continuous model delivery
CI/CD for deep checkpoints keeps athlete-facing metrics aligned with the latest validated training runs.