For Swim Tech Startups and Elite Academies. We build custom Computer Vision Engines that track stroke mechanics, turn times, and velocity in real-time—solving the complex challenges of water refraction and splash occlusion.
Tracking a human in a pool is one of the hardest challenges in Computer Vision. Off-the-shelf pose estimation models (like standard MediaPipe) fail immediately due to three physics blockers:
.png?width=2048&quality=90&format=auto)
Light bends when it hits water. A standard model thinks the swimmer's arm is 6 inches away from where it actually is. We build Refraction-Correction Algorithms to map true anatomical position.
.png?width=2048&quality=90&format=auto)
Bubbles and whitewash hide the swimmer's body. Our Temporal Logic Models predict limb position based on stroke cadence, filling in the gaps when the camera is blinded.
.png?width=2048&quality=90&format=auto){kind=tracking}
Tracking a swimmer as they break the surface (air-to-water transition) usually crashes the model. Our engines fuse above-water and underwater feeds into a single continuous biomechanical profile.
From "Smart Pool" apps to Olympic-grade performance lab
For consumer apps. Users place their phone at the end of the lane. Our AI counts laps, calculates SWOLF scores, and analyzes head position without requiring a smart watch or wearable.
.png?width=2048&quality=90&format=auto)
We calibrate the model for specific pool conditions (lighting, water clarity). We use "Background Subtraction" to isolate the swimmer from the lane ropes and tiles.
.png?width=2048&quality=90&format=auto)
We train custom models on underwater datasets. This ensures the AI recognizes a "distorted" arm as a human limb, correcting the coordinates to match physical reality.
.png?width=2048&quality=90&format=auto)
Our Logic Engine identifies the four phases of a stroke (Catch, Pull, Exit, Recovery). This allows us to calculate DPS (Distance Per Stroke) and specific phase velocity.
.png?width=2048&quality=90&format=auto)
Pools often have poor Wi-Fi. We optimize models to run offline on tablets (iPad Pro) or edge gateways so coaches get feedback instantly on the pool deck.
.png?width=2048&quality=90&format=auto)
"We wanted to build a 'Smart Gym' experience without rewiring the facility. Folio3 built a vision system that runs on standard security cameras, automatically tracking stats for every player who steps on the court." — CTO, Basketball Analytics Startup.
Whether you are building a safety tool or a performance app, we have the engineering team to execute.
Contact us at +1 408 365-4638 or email us using the form below. via contact@folio3.ai
Field Experience
Done Around World
Client Satisfaction
Established On
Response Time
+1 408 365-4638
contact@folio3.ai
6701 Koll Center Parkway, #250 Pleasanton, CA 94566
