Docs/Face Recognition

Face Recognition

PRO

Face Recognition is an advanced feature available on select PRO categories that identifies drivers and team personnel by their facial features — even when race numbers aren't visible. Currently available for IMSA WeatherTech.

How It Works

PRO categories use custom-trained ONNX models that run locally on your computer. When Face Recognition is enabled for a category, the AI performs two parallel analyses on each image:

🏎️ Vehicle Detection

Identifies the car, reads race numbers, matches livery patterns, and classifies the scene type (on-track, pit lane, podium, etc.).

🧑 Face Detection

Detects faces in the image, extracts facial features, and matches them against a known database of drivers for that championship.

The two detection systems work together. A portrait photo of a driver standing next to their car? The AI identifies them by face AND by car number, cross-referencing both for maximum confidence. A close-up helmet-off shot with no car visible? Face recognition handles it alone.

When Face Recognition Helps

Face Recognition is particularly valuable for:

  • Podium & celebration photos — drivers without helmets, away from cars
  • Pit lane portraits — driver headshots, team personnel, grid walks
  • Press conference shots — seated drivers with no car visible
  • Fan meet & greet — drivers interacting with fans in the paddock
  • Ambiguous number shots — where the number is blocked but the face is visible

Supported Categories

Face Recognition is currently available for:

CategoryFace RecognitionScene ClassifierStatus
IMSA WeatherTech✅ Yes✅ YesAvailable
F1 2026— Not yet✅ YesComing soon
MotoGP 2026— Not yet— Not yetIn development

Using Face Recognition

Face Recognition is automatic — there's no toggle to enable or disable it. When you select a PRO category that supports it (like IMSA WeatherTech), the face detection model runs alongside the vehicle detection model. Results show both recognition sources, and the AI uses the strongest signal to make its final match.

No extra token cost: Face Recognition runs locally on your machine using ONNX models — it's included in the PRO category analysis at no additional token cost. The local processing also means it works offline after the initial model download.