InteriorLAB

We're breaking past surface-level chatbots. Designers take days to make models. From 5 days to 5 SECONDS, Our app converts human-prompt to a customizable 3D model on our own visualization software.

Comment

DEMO VIDEO: https://youtu.be/dwH-ssjJKk0

INFO DECK: https://docs.google.com/presentation/d/1KWeCWUIWsASWHxQ5EM79Ij160dxvPHz9sg2TYTI31I0/edit?usp=sharing

After speaking to Roopika, an architect student, she mentioned how CAD designs take hours on end to create this “perfect” result; stating how iterating on the new implementation of certain designs involves lots of energy, causes a loss of time, and also has a lot of changes which takes a long time to produce. This concept of “constant change” and the rise of AI and chatbots made us want to develop a chatbot capable of autonomously enabling real-time, 3-D visualization of designs for furniture, beds, and other interior parts. This will allow architects to generate designs in <5 seconds, automating a process that usually takes hours on end.

“This would be game-changing” - said Roopika, an Architecture student.

What are the main features?

Through the development of this chatbot, there are many key features involved. Firstly, we start with Python code, merged with GPT API, OpenSCAD, and VTK, enabling the manipulation of code generated by the software. Merging this code with AI enables seamless integration of the desired designs into 3-D visualizations. The design is initially written in “SCAD” code, which is then converted into “STL” code, which can then be put into our terminal app called “InteriorLab.” Code is automatically placed into the terminal app, which enables the chatbot to directly convert that “ask” that the users have into real-time autonomous updates using seamless 'back-and-forth' communication.

We also created our very own 3D visualization software to complement our AI model using Python's Visualization Toolkit.

What did we use to build InteriorLAB?

We used Python to build the backend components, with the help of the ChatGPT 4 OpenAI API Key, OpenSCAD; a design visualization software, aswell as VTK; a visualization toolkit that enables image processing and scientific visualization. For the front-end components, Tkinter was used, allowing an appealing interface that the user engages with. InteriorLab acts as an app that revolutionizes the way automation occurs for designers, architects, and more!

What Challenges did we face?

Automating the entire application was a big problem that we faced, mainly because this is quite novel from existing models, which deters from automating such a niche topic as 3-D design. Furthermore, SCAD to STL was another huge problem that we faced, we had to try reworking the code to reach the goal. Lastly, putting this into VTK (the viewpoint part of our app) was also another milestone challenge that we had to overcome as a result. Overall, we are very pleased with the final result and the impact that this application can have!

What are we proud of?

We are proud of the fact that we generated something that is novel (never been done before) and that we can culminate all of the components and merge that into a self-functioning and autonomous app that can be used by the >3.2 million architects in the world, aswell as serve as a versatile app that enables people far from architects, such as 3-d designers, 3-d printers, and those in the interior space.

What did we learn?

Through the development of this application, we learned how to merge and utilize various unknown technologies, helping us overcome many loopholes along the way. We learned how to manipulate the OpenAI API key, something that seemed foreign to us, OpenSCAD, aswell as VTK, and combine all of these facets into structured code that serves an important goal in generating 3D, Interior Designs.

What is Next for InteriorLab?

We would love to generate a more appealing front end, which can be accessed from the browser. This will enable more people to utilize the app as we take this project further post-hackathon!

Built With

Share this project:

Updates