Furniture Removal Pipeline

XDD (Chenyue Dai)

Abstract

This project presents a comprehensive approach to data-driven home workspace design, leveraging advanced technologies such as 3D scanning and AI-driven algorithms. The initiative, led by Wangli (SMArchS Computation ’24) with support from xdd (MArch ’26), focuses on developing a Furniture Removal Pipeline and exploring integration with tools like PolyCam Room Mode and Revit.

The Furniture Removal Pipeline is designed to generate clean room models by identifying and removing furniture, reconstructing geometry, and applying realistic textures using clustering methods. Key functions include recognizing furniture, correcting scale, and providing detailed drawings and measurements. The pipeline's efficacy is demonstrated through the successful reconstruction of various room geometries and the extraction of dominant colors for texture application.

Challenges addressed include the limitations of LiDAR technology, the need for extensive calculations, and the difficulties in recognizing walls and curved surfaces. Future directions involve enhancing the integration with Revit for sophisticated design management and exploring generative spatial design.

Ultimately, this project aims to democratize home workspace design by making it more efficient and accessible, offering innovative solutions for clean room modeling and accurate spatial design through the use of emerging technologies.

The following is a conceptual example of how the project could be used to support user-driven design requests. In this case, a user expresses the desire to add a new garage to the side of their house. The plan view of the house's floor layout shows the existing rooms, with the desired location for the new garage highlighted in red. The red line connects the user’s request to the specific area on the floor plan where the new garage would be added, demonstrating how user input can directly inform and guide spatial design modifications. This approach highlights the project's potential to incorporate user preferences into the automated generation of spatial designs, such as adding new structures like a garage based on the user's specifications.

The project centers around the development of a data-driven home workspace design app, which considers factors such as proximity to windows and the impact on mood. A key component of the project is the Furniture Removal Pipeline, which operates under the assumptions that walls are straight and perpendicular at the corners. The process involves identifying wall directions by treating surface normal as a point cloud, removing points that are far from the XY plane, and using clustering techniques to determine the precise orientation of the walls. Additionally, the pipeline includes a color extraction process to enhance the visual accuracy of the reconstructed room models.

Use K-mean clustering to extract dominant colors for creating fake textures.

Geometry Reconstruction:

Extract the outline using algorithms.

Apply vertical extrusion to clean room geometry.

Use Delaunay triangulation for meshing.

 

The project considers integrating the pipeline with Revit to enable more sophisticated design management, offering benefits such as natural control over element positions and the ability to embed useful information directly into the model. Current progress includes constructing walls and floors from code, with future directions exploring advisory and generative spatial design through Revit. However, the approach has limitations, as it requires LiDAR technology, which is restricted to certain Apple devices and has a range limit of less than 5 meters. Additionally, walls may not be recognized if they are too close to each other, and the process demands extensive calculations, including the use of neural networks.

 

Conclusion:

The project is dedicated to enhancing the design and reconfiguration of home workspaces by harnessing the power of advanced technologies such as 3D scanning, AI-driven algorithms, and sophisticated design software. By integrating tools like PolyCam for detailed 3D room scans and Revit for advanced design management, the project aims to offer a comprehensive workflow that includes clean room modeling, efficient furniture removal, and precise spatial design. This approach not only streamlines the design process but also ensures accuracy and flexibility, allowing for tailored design solutions that meet the specific needs and preferences of users. The project’s ongoing development and future directions promise to further refine these capabilities, paving the way for innovative applications in home workspace design.