Inspiration
Designing a home is joyful, as each member of the family has their own ideas and preferences. Interior designers help clients walk through this process by showing multiple plans and different furniture arrangements. However, this process can become lengthy and repetitive.
This project was inspired by the idea of making that process simpler and helping users design the layout provided by their real estate developer while allowing them to view the design from different viewing angles.
What it does
HouseSpaceAI helps users design the layout provided by a real estate developer and visualize it from different viewing angles.
The main goals while building it were:
- Simple to use
- Faster and more accurate results
- Design the dream
How we built it
I used different approaches to solve the problem:
Image to 3D:
I first tried converting an image directly into a 3D layout, but this affected the accuracy.Image to 2D floor plan to 3D:
I then tried converting the image into a 2D floor plan first and then into 3D. This improved the accuracy by a bit.
There are still some issues with furniture orientations and sizes.
Challenges we ran into
The biggest challenge was maintaining accuracy while converting a provided layout into a usable 3D representation.
- Direct image-to-3D conversion affected accuracy.
- Adding an intermediate 2D floor-plan step improved accuracy, but it was still not perfect.
- Furniture orientation and sizing remain areas that need further improvement.
Accomplishments that we're proud of
- Built a working concept for turning a real estate-provided layout into an interactive design.
- Explored different approaches to improve the accuracy of the generated layout.
- Focused on making the experience simple and faster for users.
- Created a way to view the designed space from different angles.
What we learned
Building this project gave me experience with new libraries such as WebMCP and tools that make connections with the browser.
I also learned how different approaches to image and floor-plan processing can affect the accuracy of a 3D representation.
I still feel that not having a Codex Plus connection was a disadvantage, as it would have helped me test the product more extensively.
What's next for HouseSpaceAI
The opportunities are limitless. HouseSpaceAI can potentially be used by:
- Real estate developers
- Interior designers
- Individuals designing their own homes
The goal is to continue improving accuracy, especially furniture orientation and sizing, while making the experience faster and easier so more people can design their dream spaces.
Built With
- codex
- three.js
- vite
- webgpu
- webmcp
Log in or sign up for Devpost to join the conversation.