Inspiration
We saw that people in different cities tend to look at different apartments to see if its a perfect fit for them, to see if all of the furtinute that was in their old house would
What it does
It takes real world objects, automatically converts them into virtual and moveable objects which can then be moved around in our custom environment
How we built it
Konva (react-konva): Handles the precise 2D/Isometric drag and drop grid system and collision detection Three.js : Powers the True 3D view and renders the procedurally generated furniture models.
We integrated Google's Gemini 2.5 Flash API. When a user uploads a photo of a real object, we pass that into Gemini with specfic guidelines to describe the picture, and then we pass that along to generate an actual model, which returns executable React Three Fiber code representing that object using standard geometries
Challenges we ran into
It was really hard to get the model to spit out good models because we didnt have a good baseline
Accomplishments that we're proud of
The incorporation of Gemini to create a powerful tool to import your real world furniture and objects into your room editor via computer vision. The multitude of features that allow for a variety of viewing angles and smooth and simple movement.
What we learned
We learnt that incorporating real life customizable objects into digital form is an extremely complicated process, one that requires smart usage of computer vision and AI, as well as limiting its features so that the output is consistent.
Log in or sign up for Devpost to join the conversation.