Eliminating Manual Data Entry in Architecture Digitisation
Optimise for 3D rendering with AI-powered solutions.
A typical challenge to the construction industry is the void between the potential end-product and the present condition of a residence. Converting 2D floorplans to 3D can solve this issue as it allows businesses to envision the project’s viability, increases accuracy, and aids quick altering without wasting assets like time and budget. However, the process of generating tables of quantities, allocations, and 3D drawings requires a lot of time and man-hours.
Our client, a large Japanese construction corporation with a global presence, had their staff manually convert 2D data on the blueprint to 3D data. However, the number of people who could handle the process, as well as the time required, were limited. This led to the restriction of cases that could be processed and cost overruns.
Some of the biggest challenges this client faced were:
1. Various types of input documents
Construction entities were having issues with data extraction from complex documents like blueprints, pipeline layouts, industrial plans, manufacturing schemes and maps obtained from the third-party vendors and partners.
2. Manual tasks
To render to 3D, all information/ data from a 2D floor plan is mandatory. This was a repetitive and manual task as it required human effort to read information on the blueprint, checking for the number of rooms, widths and heights, position of wall, door, window, etc.
Once the data in the 2D floor plan was acquired, the data had to be entered into an internal system, which was both time consuming and required expert skills.
To solve this problem, we proposed to use AI models to read the 2D blueprints, extracting required information so that it can be used as input for the 3D rendering.
Data were extracted from a 2D floorplan by:
- Creating AI models extracting data such as baseline position and name, room compartment, name of the room, position of windows and doors in 2D, windows and door height, position of walls in 2D, position and height of roof surface, etc.
- Testing at least one extracted data with the recognition response rate over 70%
- Verifying automatically against data sets, building material selection and layout work from coordinate data.
As a result, our solution was able to eliminate the manual tasks involved in data-entry, boost efficiency and decrease costs.