When a 3D model of the object to be placed (top left) and the placement instructions written in text (bottom left) are input, a space with the objects arranged according to the instructions (right) is generated as output.
Abstract
We propose a system that automatically arranges 3D models based on text instructions. Users create 3D models of the objects they want to place using 3D modeling software and input them into a virtual environment. Next, they enter placement instructions in text, such as "arrange in a circle." The system then generates a prompt incorporating these instructions and sends it to ChatGPT, an LLM. In response, the system retrieves the 3D model's coordinates and rotation and reflects them in the virtual environment, enabling automatic arrangement. By using this system, users can simulate object placement in a virtual environment without physically moving objects in the real world.
BibTeX citation
@Misc{nakase.seiji.21_bachelor,
author = {Seiji Nakase},
title = {Object Layout Proposal System Using Text Instructions with LLM},
howpublished = {Bachelor Thesis},
month = 2,
year = 2025,
}
