8 February 2024
Computer Vision and Pattern Recognition
Enables editing driving scene simulations via language
Uses collaborative LLM agents for flexible editing
Employs novel multi-camera neural radiance field method
Proposes lighting estimation to integrate assets
Shows complex command handling and photo-realistic results
Editable simulation of autonomous driving scenes
This paper introduces ChatSim, a system that enables editing photo-realistic 3D driving scene simulations through natural language commands. ChatSim uses a framework of collaborative large language model agents to interpret commands and distribute specialized editing tasks. To generate high-quality outcomes, ChatSim employs novel methods for multi-camera neural radiance field rendering and lighting estimation with external digital assets.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper