Researchers at Carnegie Mellon University have introduced 'LegoGPT', an innovative AI model that generates Lego designs from text prompts. This tool allows users to describe any object, such as 'an acoustic guitar with an hourglass shape', and LegoGPT will determine the pieces and assembly instructions required. Currently, the model supports 20 categories, with plans for expansion. Built on Meta's LLaMA-3.2-Instruct-1B, LegoGPT was trained on a significant database of Lego structures and 3D shapes and may have broader applications in architecture and design.
The LegoGPT demo allows Lego enthusiasts to transform text prompts into buildable designs, generating necessary pieces and assembly instructions based on trained AI models.
Researchers trained LegoGPT on a database of 47,000 Lego structures and 28,000 unique 3D shapes, enabling it to predict designs from simple textual inputs.
Collection
[
|
...
]