Adaptive multimodal large language model with hierarchical cross modal grounding for real time industrial visual anomaly detection on edge devices
The HCG-MLLMAD - Hierarchical Cross-Modal Grounding framework of Multimodal Large Language Model-based Anomaly Detection - a variant of the original Light-MLLMAD architecture with multi-scale visual feature pyramids, prompt routing, and a hierarchical cross-attention fusion mechanism is presented.
P. Sundaravadivel, U. Kumaran, E. S. Vinoth Kumar et al.
· Discover Artificial Intellig... · 0 citations