Skip to main content

Race Number Recognition

Development of an automated race number recognition system for motorcycle racing – end-to-end inference pipeline with YOLO-based detection, OCR fine-tuning, and .NET integration.

Client/Company/Industry

Christian Engelhardt Softwareentwicklung

Duration

6 months

Product

Software

Expertise

Software Development

Goal

A motorsport software developer needed an automated pipeline for detecting race numbers on motorcycles. The solution had to work reliably under real racing conditions - with variable lighting, motion blur, and diverse typefaces on number plates - and integrate into an existing .NET application on Windows.

Tasks

  • Data labelling for race number plate detection and OCR training
  • Training YOLO models for race number plate localisation
  • Fine-tuning OCR models for accurate digit recognition
  • Development of an end-to-end inference pipeline combining detection and OCR
  • Benchmarking models for CPU-only deployment scenarios
  • Integration of the pipeline into a .NET application via PythonNET (C#, Windows)

Challenges

The main challenge was the variability of input data: race numbers on motorcycles appear in different sizes, angles, typefaces, and lighting conditions. Models had to be robust against this variability without relying on specialised GPU hardware.

Fine-tuning OCR models on motorsport-specific typefaces and image conditions required careful data curation and iterative evaluation to ensure reliable digit recognition in real-world use.

Programming Languages

Python, C#

Technologies

YOLO, TensorRT, ONNX, OpenCV, PythonNET

Project Image

Race number recognition inference pipeline showing detection bounding box and OCR output.

Similar problem?

Contact us

Takeaway

The result is a complete end-to-end race number recognition pipeline that runs stably on CPU and integrates seamlessly into the client's existing .NET application. The combination of YOLO detection and a fine-tuned OCR model delivers reliable results under real race conditions.

Similar Projects

Project Image

RIM2D - Highly Efficient 2D Hydraulic Simulation of Fluvial, Pluvial, and Urban Flooding

Hydrodynamic Simulation Web Application Geodata GPU Computing

RIM2D is an existing, highly efficient 2D hydraulic simulation model for fluvial, pluvial, and urban flooding. As part of a strategic partnership, we supported the extension of the research code with a web application and a cloud-based GPU simulation environment, enabling its transition into a market-ready product.

Project Image

Data Lake for Geoscience Research Data Management

Datalake Research Data Management Geodata Cloud-Native Open-Source

We developed an Open-Source S3-based data lake solution for the centralized ingestion, categorization, and searchability of data. The goal was to automate and improve manual data management through an integrated architecture with workflow orchestration, data cataloging, and access control.

Back To Top