Computer Vision

Real-Time Ball Tracking

TruPutt employs OAK-D Lite depth cameras paired with custom-trained YOLOv8 neural networks. Our detection algorithms identify the ball position with sub-millimeter precision, while Kalman filtering ensures smooth, jitter-free tracking even at high speeds.

OAK-D Lite YOLOv8 Kalman Filter 60 FPS
Physics Engine

Templeton-Penner Model

Our physics engine implements the professional-grade Templeton-Penner putting model, accounting for rolling resistance, surface friction, gravity vectors, and slope effects. Combined with AimPoint algorithms, predictions match tour-level accuracy.

Templeton-Penner AimPoint Auto Stimpmeter Gravity Vector
v(t) = v0 - (mu * g * cos(theta)) * t
Surface Analysis

LiDAR Height Mapping

iPhone LiDAR scans capture your putting surface in extraordinary detail. RANSAC plane fitting algorithms extract true undulations, generating a precise 3D height map that reveals every subtle break invisible to the naked eye.

LiDAR Point Cloud RANSAC Fitting Grid Interpolation Break Analysis
Visual Feedback

Projection Mapping

Real-time visual feedback is projected directly onto your putting surface. Homography transformations ensure pixel-perfect alignment between camera coordinates and projector output, displaying aim lines, predicted paths, and game graphics exactly where they belong.

Homography Camera-Projector Calibration Real-Time Rendering
System Architecture

Data Flow Pipeline

Capture

OAK-D camera captures depth + RGB at 60fps

Process

DepthAI pipeline runs on-device neural inference

Analyze

Physics engine predicts trajectory from tracking data

Render

Projection maps graphics onto putting surface