Transformer Floor Plan Generation From Multi-Room Building Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately and efficiently generate building floor plans from visual data without depth sensors, particularly in multi-room environments, due to challenges in capturing, representing, and using building interior information effectively.
Innovation Solution
A combination of a trained diffusion transformer machine learning model and a bundle adjustment optimizer model is used to analyze visual data from multiple images, determining global inter-image pose and wall locations to generate precise floor plans, even with limited image overlap, by leveraging geometric priors and denoising diffusion architectures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional floor plan construction methods are used, then accuracy of building interior information can be maintained, but the complexity and difficulty of construction and maintenance increases significantly
Solution Approach 1:
The patent replaces manual mechanical floor plan construction with an automated computer-based system that processes images and generates floor plans algorithmically. The system uses image processing and computational geometry to automatically extract wall locations, room boundaries, and structural elements, eliminating the need for manual measurement and drafting while maintaining high accuracy.
Solution Approach 2:
The system creates digital copies of building interiors by processing photographs and generating standardized floor plan representations. These digital floor plans serve as accurate replicas of the physical space, enabling remote viewing and analysis without requiring physical presence or manual reconstruction.
2Loss of information
If manual floor plan creation is used, then detailed building information can be captured, but the time required for construction and updates increases
Solution Approach 1:
The automated system rapidly processes multiple images and generates comprehensive floor plans in minutes rather than hours or days. The computational process simultaneously extracts multiple types of information (walls, doors, windows, rooms) from images, maintaining completeness while dramatically reducing construction time.
Solution Approach 2:
The system performs preliminary automated analysis of building images to extract all necessary structural information before final floor plan generation. By pre-processing images to identify and locate all building elements, the system prepares complete data sets that can be quickly assembled into accurate floor plans without time-consuming manual work.
3Ease of operation
If remote building inspection is implemented, then physical travel is eliminated, but the ability to accurately capture and represent building interior information becomes difficult
Solution Approach 1:
The system replaces physical inspection with automated image processing that accurately captures building interior information. By using computer vision algorithms to analyze photographs, the system extracts precise measurements and structural details remotely, maintaining information accuracy without requiring physical presence in the building.
Solution Approach 2:
The system uses images as an intermediary between the remote inspector and the building interior. These images serve as the medium through which building information is captured, transmitted, and processed, enabling accurate remote representation of the physical space without direct physical access.
4Loss of information
If floor plans are used for navigation, then building layout information is provided, but the visualization and usability for navigation purposes is limited
Solution Approach 1:
The generated floor plans serve multiple functions beyond traditional static representation. They are designed to support various navigation tasks including pathfinding, location identification, and spatial orientation. The standardized format and accurate geometric representation enable the same floor plan to be used for multiple navigation purposes and integrated with different navigation systems.
Data Source
AI summary
Techniques are described for automated operations to analyze visual data from images acquired in multiple rooms of a building to generate building information that may include a floor plan for the building, such as by analyzing visual overlap between those images to determine information that includes image pose data for the images and wall location data for walls of the rooms that are visible in the images, and by using the generated building information in further automated manners. In some situations, the described techniques include using a trained transformer machine learning model to encode and compare information from some or all pixel columns of the images to map pixel columns to particular walls, and to use that data along with floor-wall boundaries in the pixel columns to generate a resulting floor plan for the building.