Objective rehabilitation,
measured at a distance.
Built with clinicians, grounded in clinical practice.
Reference pipeline reports 91.7% side-view and 100% front-view correct exercise identification on the knee PoC dataset (arXiv 2412.20733).
Why RehabVision
Real-time biofeedback
Markerless pose estimation runs directly in the browser at clinical frame rates, with end-to-end latency below the 100 ms human perception threshold.
Consortium partners
At a glance
Correct exercise ID — side view
Per the arXiv 2412.20733 PoC reference pipeline.
Correct exercise ID — front view
Same reference pipeline; target for RehabVision pilot.
Clinician time reduction target
BG-02 — clinician minutes per supervised exercise.
Test suite coverage
Pure-module unit tests covering CSV, build and utility code.
Product preview
See the workspace in action
Dashboard
RehabVision workspace
Why RehabVision
Six benefits the project is contractually committed to deliver
Each benefit is traced from the project business case (Benefits B1–B6 in the RehabVision Project Plan v0.3) and reports against measurable success metrics at project closure.
More accurate rehabilitation
AI assessment of exercise execution quality with automated error detection flags incorrect movement at the frame level rather than waiting for the next clinic visit.
Scalable availability
A thin web client over a REST backend means no specialist hardware at the clinic or the patient's home — a modern smartphone or webcam is enough.
Objective patient progress
Automatic range-of-motion, correctness and pain tracking replaces clinical memory with a longitudinal, timestamped record the clinician can act on.
Scientific validation
A peer-reviewed publication with methodology and results, co-authored with AUT and the project coordinator FT Park z.ú., grounds the platform in independently verifiable evidence.
Commercial sustainability
A SaaS licensing model for rehabilitation facilities, defined after the clinical pilot, ensures the platform continues to serve patients beyond the initial research phase.
Interoperable by design
Standard CSV and JSON exports, FHIR-compatible session metrics, and an OpenAPI backend let the platform plug into existing clinical IT — no proprietary lock-in for the clinic or the research partner.
The method
Movement analysis, grounded in clinical practice
RehabVision does not replace the clinician. It replaces the blind spot between visits — the six days a week when a patient's recovery used to be invisible. The knee PoC pipeline is deliberately simple, privacy-preserving, and auditable.
Capture
The patient performs a prescribed exercise in front of any modern smartphone or webcam. MP4 input at 25–60 FPS. No markers, no wearables, no calibration.
Capture
The patient performs a prescribed exercise in front of any modern smartphone or webcam. MP4 input at 25–60 FPS. No markers, no wearables, no calibration.
Estimate
MediaPipe Pose extracts anatomical keypoints frame by frame, converting raw video into a privacy-preserving time series of joint positions — no faces, no PII, only a stick-figure overlay.
Estimate
MediaPipe Pose extracts anatomical keypoints frame by frame, converting raw video into a privacy-preserving time series of joint positions — no faces, no PII, only a stick-figure overlay.
Analyse
Movement-specific heuristics compute knee angles, range of motion, repetition count and frame-level exercise correctness — the same metrics a clinician would assess in person, at roughly 30 samples per second.
Analyse
Movement-specific heuristics compute knee angles, range of motion, repetition count and frame-level exercise correctness — the same metrics a clinician would assess in person, at roughly 30 samples per second.
Report
Every session becomes a review-ready record: charts, raw CSV in schema v0.1, and a replayable stick-figure overlay. The clinician can intervene early instead of reacting late.
Report
Every session becomes a review-ready record: charts, raw CSV in schema v0.1, and a replayable stick-figure overlay. The clinician can intervene early instead of reacting late.
Clinical voices
What our pilot partners say
“The frame-level correctness detection caught movement compensations we'd been missing during in-person visits. It changed how we prescribe home exercises.”
Dr Eva Králová
Physiotherapist · FT Park z.ú.
Built with clinicians and patients
Two primary users, one shared record of truth
Pricing
Simple, transparent pricing
Start free during the pilot. Upgrade when your clinic needs multi-patient management and clinical reporting.
Pilot
For individual patients during the clinical pilot phase.
- 1 patient workspace
- Unlimited sessions
- ROM + correctness tracking
- Pain self-reporting
- Progress charts
- CSV export
Clinic
For physiotherapy practices with up to 20 patients.
- Up to 20 patients
- Clinician dashboard
- Patient assignments
- Clinical reports
- Audit trail
- Email support
Enterprise
For hospital networks and research institutions.
- Unlimited patients
- Multi-site management
- API access
- SSO integration
- TWIST compliance pack
- Dedicated account manager
FAQ
Frequently asked questions
Start measuring what you used to estimate
A free account gives you access to a personal workspace, guided exercises, and your own progress history. Upgrade when your team needs multi-patient management and clinical reporting.