.png){kind=icon}
.png){kind=icon}
Return
LLM-centric collaborative computing framework: Leveraging industry specified knowledge for open-set visual recognition
.png){kind=icon}
PRE
AI
DOI：10.1016/j.asoc.2025.114028.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A LLM-centric device-edge-cloud architecture is proposed for industrial visual recognition, which achieves efficient open-set recognition results with rational allocation of computation burden brought by LLMs. • A multi-server microservice deployment strategy is proposed, which offers deployment solution of multiple servers to balance computational cost and achieve minimal delays. • A LLM collaboration scheme is proposed, which leverages cycling workflow of LLMs and knowledge graph to generate task-specified knowledge, thus improving robustness and accuracy to recognize unseen objects.