About TyreCam AI
Automated optical tire tread depth analysis & physical safety scoping
Our Mission
TyreCam AI (an evolution of the ScanTire AI project) is an automated computer vision system designed for non-destructive, contactless inspection of pneumatic tires. We provide instant tire AI scanning and tread depth estimation, Wear Indicator (TWI) verification, and diagnostic PDF reports for car owners, used car inspectors, auto repair shops, and commercial fleets worldwide.
The Computer Vision Challenge
Optical Tread Geometry Analysis
Analyzing tire tread from a single 2D camera photo presents significant challenges: variable shadow depth, rubber micro-textures, dirt distortion, and angled perspectives. TyreCam solves this by combining client-side defect gating with a multi-model server-side neural pipeline (deep learning) and a Physics-Based Fusion Engine.
Our Technology Architecture
Stage 0 Gatekeeper
In-browser ONNX model pre-filters non-tire images and flags critical physical defects (cuts, hernia, cord exposure) before server analysis.
Multi-Model Depth Pipeline
Server-side ONNX models run in parallel: ROI detection, Depth U-Net, 5-class wear classifier (ResNet architecture), and geometric TWI grid/heatmap consensus.
Physics-Based Fusion Engine
Fuses neural predictions with physical tire constraints, applying Safety Shield bracket clamping and conservative range estimation.
ONNX Runtime Core
Optimized CPU-threaded inference pipeline built on ONNX Runtime for low-latency, high-throughput image processing and computer vision mathematics.
System Capabilities & Roadmap
PDF Diagnostic Reports
Downloadable summary with metadata for digital archives.
Sidewall OCR (Roadmap)
Planned feature: automated embossed size extraction.
Current Production Status
- • Live hybrid AI pipeline with Safety Shield guardrails.
- • Optimized for global standards (EU 1.6mm / US 2/32" legal minimums).
- • Continuous neural dataset expansion and accuracy improvements.