Cutting Floor Plan Digitization Time by 80% for Horizon Spaces
Client
Horizon Spaces
Country
Norway
Industry
PropTech
Team Size
2
Service Used
Engagement Length
Dec 2023 – July 2025
About Client
Turn operational data into useful insight. We bring together dependable pipelines, clear reporting and accessible interfaces so teams can make informed decisions.
From initial discovery to ongoing support, Prime Horizon Technologies brings structure and care to software delivery. We translate business needs into maintainable applications designed to evolve.
Project Overview
Many real estate professionals struggle with manual floor plan digitization. This task takes hours and often leads to errors. Horizon Spaces saw an opportunity to automate this using modern technology. They wanted to convert simple images into editable digital files.
However, they needed deep technical expertise to build a production-ready system. They approached Prime Horizon to turn this vision into a scalable reality.
Prime Horizon provided specialized engineers to build the core engine and infrastructure. The team developed a sophisticated pipeline that converts images into structured data. This solution allows users to manage architectural drawings with ease.
Today, Horizon Spaces offers a seamless digital experience to its clients. The partnership has turned a complex manual process into a fast, automated service.
However, they needed deep technical expertise to build a production-ready system. They approached Prime Horizon to turn this vision into a scalable reality.
Prime Horizon provided specialized engineers to build the core engine and infrastructure. The team developed a sophisticated pipeline that converts images into structured data. This solution allows users to manage architectural drawings with ease.
Today, Horizon Spaces offers a seamless digital experience to its clients. The partnership has turned a complex manual process into a fast, automated service.
Technologies Used
Python
PyTorch
MMDetection
MM CV
OpenCV
PostgreSQL
GCP
Vertex AI
Cloud Run
Cloud Storage
DVC
Docker
GitHub Actions
Challenges faced by the Client
Horizon Spaces needed to solve several complex technical hurdles to make their platform viable for professional architectural use. These problems ranged from handling inconsistent data sources to managing the heavy computational needs of a growing user base. The following points describe the specific obstacles that slowed down their initial development.
1. Inconsistent Drawing Standards
Architectural plans vary significantly in visual style, scale, and image resolution. This inconsistency made it difficult for the software to identify structural elements accurately across different document types.
2. Structural Integrity Issues
Objects like walls and doors often appeared as disconnected segments after the initial detection. Because these elements were not logically linked, the resulting digital maps were often unusable for professional editing.
3. Development Bottlenecks
The team originally managed datasets and model versions through manual processes. These repetitive tasks created a bottleneck that prevented the company from releasing frequent updates to their technology.
4. Infrastructure Limitations
Processing high-resolution images requires a robust and scalable cloud environment. The client lacked a production-ready system that could handle multiple inference requests without experiencing lag or downtime.
Solutions by Prime Horizon
Prime Horizon addressed these issues by designing a custom artificial intelligence ecosystem and a modern engineering infrastructure. The team focused on creating a seamless flow between raw data ingestion and the final digital output. Below are the specific engineering strategies Prime Horizon used to build a more reliable platform.
1. End-to-End Automation Pipeline
Prime Horizon built a comprehensive pipeline that handles every stage of the digitization process. This system moves an image through preprocessing and object detection to produce a structured digital file automatically.
2. Spatial Reasoning Layer
The team implemented a specialized logic layer to help the system understand how different objects relate to one another. This ensures that every detected wall or window connects perfectly to form a realistic room layout.
3. Production MLOps Framework
Prime Horizon established an automated MLOps workflow to track experiments and manage model versions. This framework allows engineers to retrain and deploy improved models without interrupting the live service.
4. Scalable Cloud Architecture
The engineers developed a cloud-native environment on Google Cloud to separate model training from user requests. This separation ensures the platform remains responsive even when the system is processing large datasets.
Outcomes of the project
From initial discovery to ongoing support, Prime Horizon Technologies brings structure and care to software delivery. We translate business needs into maintainable applications designed to evolve.
Drastic Reduction in Manual Labor
From initial discovery to ongoing support, Prime Horizon Technologies brings structure and care to software delivery. We translate business needs into maintainable applications designed to evolve.
Improved System Reliability
Prime Horizon’s cloud infrastructure enabled the platform to scale seamlessly as the user base grew. As a result, the system maintains high performance even during peak hours of heavy user activity.
Continuous Product Growth
The automated development cycle created a path for constant software improvement. This capability allows Horizon Spaces to refine their detection accuracy every time new architectural data becomes available.
Professional Accuracy
The introduction of spatial reasoning ensured that every generated plan meets architectural standards. This high level of precision has made the platform a trusted tool for property managers across Norway.
Team Involvement
| Resources | Count |
|---|---|
| ML Engineer | 1 |
| MLOps Engineer | 1 |
Core Features of the Software
The software provides a set of high-precision tools that bridge the gap between static paper plans and modern digital formats. These features offer the accuracy and flexibility required by architects and property managers. The following sections detail the core capabilities Prime Horizon integrated into the Horizon Spaces platform.
Automated Element Detection
The core engine identifies walls, windows, and doors within seconds of an image being uploaded. It uses advanced classification to distinguish between different structural components even in low-quality scans.
Vector Floor Plan Generation
The platform converts detected images into high-quality vector files. These files allow users to adjust dimensions and move structural lines as if they were working in a native CAD environment.
Automated Version Control
The system includes a central registry that tracks every update made to the detection models. This ensures that the platform always uses the most accurate version of the technology for every client request.
Interactive Web Editor
A built-in editing tool allows users to modify their digital layouts directly in a web browser. This interface eliminates the need for expensive third-party design software while maintaining high professional standards.
Development Timeline
Project Start Time
Dec 2023
1
Project End time
July 2025
2
Collaborative software engineering, shaped around your priorities.
Modernize Your Property Data
Prime Horizon specializes in solving complex automation problems for the real estate industry. We can help you turn manual processes into scalable digital solutions. Contact us to modernize your property software and improve your operational speed.
Tech Stack
Full
