Digitizing Floor Plan Layouts Using Machine Learning and Computer Vision
The lack of accurate floor plan information for existing buildings hinders informed decision-making for improvements and maintenance. Additionally, the need for layout changes in upcoming construction projects leads to significant time and resource losses. This project aims to address these issues by digitizing floor plans, identifying elements, and training computer vision models to facilitate efficient decision-making and reduce costs.
The problem statement
Existing buildings frequently lack accurate information on their floor plans for a variety of reasons, including their antiquity, the loss of documents, or a failure to communicate clearly throughout the process of transferring ownership. If building owners do not have access to these floor plans, they will be unable to make educated decisions on improvements, retrofits, and maintenance to ensure code compliance. This may result in additional costs and delays. The generation of point clouds for the inside of a structure is now possible because of technology that is both commonplace and very inexpensive, such as cell phones and Lidar scanners. Reconstructing a floor plan involves a number of steps, one of which is the segmentation of its many elements, such as windows, doors, and so on. Another problem statement lies in the area of buildings that are yet to be constructed as well. Many times it happens that layouts need to be changed later on because of a variety of reasons, resulting in a substantial loss of time and resources. The current work will focus on the above problem statements and provide relevant solutions.
The project objectives
Arealize currently possesses a partial dataset that will be shared exclusively with chosen developers. Throughout the project, the collected data will be incorporated into this dataset.
Project Objectives:
- Digitize floor plans based on architectural blueprints.
- Identify various types of elements/symbols within 2D images of floor plans for a building.
Project Scope:
- Collect and annotate diverse types of floor plans.
- Establish classification categories for different elements (e.g., doors, windows) found in floor plans.
- Train advanced computer vision models to detect the various elements present in the floor plans (e.g., doors, windows, walls).
- Determine relationships between each element depicted in the blueprint (e.g., a door connected to a wall).
First Omdena Project?
Join the Omdena community to make a real-world impact and develop your career
Build a global network and get mentoring support
Earn money through paid gigs and access many more opportunities
Eligibility to join an Omdena Top Talent project
Finished at least one AI Innovation Challenge
Received a recommendation from the Omdena Core Team Member/ Project Owner (PO) is a plus
Skill requirements
Good English
Machine Learning Engineer
Experience working with Smartphone Sensor Data is a plus.
This project is hosted with our friends at
Application Form
Become an Omdena Collaborator