Homu.sg

Revolutionizing Interior Design using Computer Vision Techniques and Generative AI.

Homu is a renovation web app that equips homeowners with the necessary tools to design, understand and engage quality contractors for their homes.

Find out more on our website at Homu.sg

Introducing Homu.sg

The home renovation process is expensive, inflexible, and homeowners are not satisfied with the quality of renovations as they can be poor design communicators. With generative AI, we can re-think home-building. We use generative AI to push the limits on how we design homes while integrating AI into the process so renovating a home is as easy as using software tools like Canva.

Team members

Raphael Yee Jun-Kai (CSD), Bryce Goh Ying Ming (CSD), Phua Yen Ching (DAI), Alexander Koh (ESD), Mah Yi Da (CSD)

Instructors:

  • Javier Gomez Fernandez

Writing Instructors:

  • Delfinn Tan Sweimay

Product Overview

Step 1. Select up to 6 images that describe your style. Step 2. Our AI will analyse your style preferences and generate an interior personalised to you. Step 3. Our AI takes your generated design and estimates the renovation cost for you.
/

Tech Break down

Throughout our project, we’ve always embraced bleeding-edge technology. The technical backbone of our features is also improved frequently.

Understanding Styles
Generate Designs with AI
AI Reno Cost Estimation

Our team wanted a way to break this notion of “styles” as static categories, and try to deepen our understanding of what homeowners may be seeing when they select images that they like. Utilizing a semi-supervised method with GPT-4 Vision and manual categorization, we crafted feature dimensions reflecting human style distinctions (e.g., luxury vs. natural). Iteratively adjusting dimension weightings with Gaussian mixture clustering, we honed our model for design differentiation.

 

To generate home designs, we first understand homeowners’ design preferences by decomposing images into elements like color and material, using segmentation and material classification. Then, we comprehend how these elements combine to create a space using LLMs.

Finally, we explored generating designs into constraints so homeowners can envision the design in their space. This required us to go beyond the state of the art.

To help homeowners understand their renovation cost, we employed LLMs to pinpoint built-ins and materials affecting costs. We then estimated these costs using data from our partner contractors.

 

What we have achieved

To date, we have over 400 registered users and we have successfully generated over 300 design reports. Our technology has demonstrated a capacity to reduce renovation costs by 20%, giving homeowners significant cost savings. Additionally, our project received coverage by major media outlets, including The Straits Times and Lian He Zao Bao, showcasing our work to a wider audience.

Supported by :

Acknowledgements

We extend our heartfelt thanks to our partner contractor, Lai Yew, for his invaluable support, time, and guidance throughout this project. Our gratitude also goes out to everyone who believed in us, providing their time and feedback to help shape our work.

Menu

ornament-menu

Contact the Capstone Office :

+65 6499 4076

8 Somapah Road Singapore 487372

Please fill in your information below and feedback

Would you like to play the audio?

Contact the Capstone Office :

8 Somapah Road Singapore 487372

8 Somapah Road Singapore
487372

Welcome back!

Log in to your existing account.

Contact the Capstone Office :

+65 6499 4076

8 Somapah Road Singapore 487372

Welcome back!

Log in to your existing account.