Doug Labs
Pose Estimation
Speech Recognition
The solution assists medical practitioners with neurological patient diagnostics; it automates, records, and secures neurological test results within an iPhone app
Cases
Postround
AI-Powered Mobile App for Automatic Golf Scorecard Recognition
Key results
93%Marker-based detection accuracy after adaptive tuning
90%Reduction in OCR errors through image-quality filtering
- Cross-platform app with real-time table extraction and validation
Services- Mobile DevelopmentComputer Vision
Industry- Sports
Technologies Used- React NativeOCRC++SwiftKotlin
Team- 1 AI Engineer1 React Native Developer1 iOS Developer 1 Android Developer
Duration- 4 months
Location- United States
It-Jim helped us bring our golf analytics app to production, improving OCR accuracy and delivering core image processing modules. They adapted seamlessly to changing requirements and showed a strong commitment to quality. Their flexibility and technical expertise made them a reliable partner throughout the launch.
The Challenges & Goals
Postround wanted to transform traditional golf scorecards into digital performance data. The task required solving several computer vision and ORC challenges under real-world conditions.
- 01
Accurate scorecard detection and alignment
Developing a robust CV pipeline to detect golf scorecards from handheld photos, align them geometrically, and isolate the data table despite variations in lighting, angle, and background.
- 02
Reliable text recognition and validation
Improving OCR accuracy for small, handwritten or printed text across diverse templates and handwriting styles, while ensuring correct mapping of values such as Lie, Par, and Score.
- 03
Handling variability in scorecard templates
Adapting to layout differences between scorecard designs by introducing marker-based identification and flexible geometrical transformation logic.
Solution Overview
It-Jim developed a mobile solution that combines computer vision, Optical Character Recognition (OCR), and backend integration to automate golf scorecard processing. The system detects scorecards in captured images, extracts the table area, validates its geometry, and prepares clean inputs for OCR. All core processing runs directly on the device through native Swift and Kolin modules integrated with custom computer vision. The final application works fully offline, ensuring stable performance and accurate results across different camera types, lighting conditions, and scorecard layouts.
- Stage 01
Application Foundation
The first stage of cooperation with Postround focused on rebuilding a reliable foundation for the mobile application. Within the React Native framework, the team reimplemented the essential product logic and user flows, ensuring consistent performance and design across iOS and Android.
Key steps:
- User authorization and profile management
- Game session setup and tracking
- Statistics and performance visualization
- Local data storage and synchronization with the backend
The milestone delivered a stable cross-platform base, ready for the integration of AI-driven scorecard recognition and advanced analytics.
- Stage 02
AI Pipeline Research and Design
In parallel with mobile development, the AI team focused on defining the core scorecard processing architecture. We needed to determine how to detect, extract, and interpret data from golf scorecards reliably under real-world conditions.
Key deliverables:
- Research and compared multiple approaches to blur, contrast, and marker detection
- Designed a marker-based system for precise scorecard localization and ID assignment
- Defined the geometry validation logic to ensure accurate table extraction
- Established the technical stack and integration plan for offline operation on mobile devices
This stage resulted in a validated processing pipeline design and clear specifications for future on-device AI implementation.
- Stage 03
On-Device AI Integration
After the research phase, the team focused on embedding the designed computer vision pipeline into the mobile application. The objective was to enable real-time scorecard detection, table extraction, and text recognition directly on iOS and Android devices.
What we did to meet the requirements:
- Integrated the C++ based CV core with native Swift and Kotlin modules
- Implemented the ARUCO market detection and geometric transformation logic
- Connected OCR and validation modules to the mobile workflow
- Optimized performance to maintain accuracy and speed in offline mode
The team delivered a functional AI engine running entirely on the device, enabling golfers to capture and process scorecards without internet connectivity.
- Stage 04
Optimization & Launch
The final stage focused on improving recognition accuracy, refining OCR results, and preparing the product for public release. The team optimized the AI and application layers to ensure stability and reliable performance in real-world use.
Main challenges solved:
- Tune ARUCO detection parameters to improve marker accuracy from early test results
- Enhanced OCR reliability through two-stage image quality filtering and logic validation for Lies and Par fields
- Conducted field testing on different devices and lighting conditions
- Delivered production build for App Store and Google Play release
We delivered a production-ready mobile application that met all functional and performance requirements and helped golfers digitize and analyze their game data.
- Outcome
Final Outcome
Postround now has a production-ready mobile app that turns paper scorecards into digital insights in seconds. The system runs fully offline, detects both printed and handwritten text, and automatically structures player data for performance analysis.
Key achievements:
- Released to the App Store and Google Play
- 93% marker-based detection accuracy after adaptive tuning
- Up to 90% fewer OCR errors thanks to image-quality filtering and logic validation
- Stable cross-platform MVP with integrated backend and analytics
The final product combines computer vision, OCR, and mobile engineering in a single, reliable solution that meets both technical and user expectations.
Let’s Talk About Building AI That Works in the Real World
No vague promises. No pressure. Just a real conversation with our technical leadership to understand your needs and explore the best next steps.
What Happens Next?
- Your message goes straight to our core team – not a sales department.
- You’ll talk directly with our CEO, CTO, or a domain expert for a consultation-style first call.
- We’ll ask the right technical questions to understand your challenge and scope.
- If we’re a fit, we’ll follow up with a proposed path forward.
More Projects Like This
Pardigm
Computer Vision
Edge AI
Hormone tracking application powered by Computer Vision with 93%+ accuracy in real-life conditions.