BUSINESS CHALLENGES
Our Client
DIGI-TEXX’s client is a software company based in South Korea, specializing in Spatial Digital Twin technology. It primarily serves the Architecture, Engineering, and Construction industries, as well as Facility Management.
The client focuses on transforming real-world spatial data into accurate digital environments to optimize building design, operation, and maintenance.
Project Challenges
The client needed to improve the accuracy and synchronization between 3D-captured building data and 2D architectural drawings. However, the project posed multiple challenges:
- Handling high data volume efficiently – thousands of images and videos daily – while meeting strict project deadlines.
- Complex spatial matching between different data formats (2D drawings vs. 3D visual captures).
- Maintaining annotation precision and consistency across large datasets.
Unclear or inconsistent student handwriting further complicates the grading process
Project Scope
DIGI-TEXX provided a dedicated annotation team working directly on the client’s proprietary system to:
This solution ensures high accuracy even in cases with erased or modified answers, so that the final scores accurately reflect the student’s performance.
- Review captured building images and videos.
- Identify and annotate matching anchor points between captured visuals and floor plans.
- Perform 2D and 3D mapping annotation based on project-specific BIM guidelines.
- Ensure annotation outputs are aligned with spatial modeling standards to reduce errors in digital twin representations.
SOLUTION
DIGI-TEXX applied a hybrid human-in-the-loop data annotation approach, combining skilled human annotators with AI-assisted tools. The team:
- Implemented rigorous quality control processes to maintain consistent annotation accuracy across all outputs.
- Created and refined anchor point matches between captured imagery and architectural drawings.
- Aligned datasets to minimize spatial gaps and improve geometric accuracy.
Our process ensured precision, scalability, and transparency across every project phase. Project Flow:
- Onboarding & Training: The team was trained on BIM-specific annotation guidelines and the client’s proprietary tools.
- Data Intake: Receive and categorize captured building images and videos from the client.
- Annotation Execution: Create matching anchor points between real-world captures and architectural floor plans (2D and 3D mapping).
- Alignment & Verification: Gradually align both data types to minimize spatial discrepancies.
- Quality Assurance (QA): Apply multi-level review to ensure consistent annotation accuracy.
- Reporting & Feedback: Submit annotated outputs, track daily progress, and incorporate client feedback for continuous improvement.
BUSINESS OUTCOME
- 20,000 items are annotated monthly, ensuring a steady and reliable data flow for model development.
- 1,000 videos processed per day, demonstrating high productivity and scalable operations.
- 95% annotation accuracy, guaranteeing precise alignment between real-world visuals and digital floor plans.
