Multimodal large language models for food safety detection within deep learning frameworks: a review.

Area of Science:

• Food Science
• Artificial Intelligence
• Supply Chain Management

Background:

• Food safety faces increasing complexity due to evolving industries and global trade.
• Traditional detection methods (e.g., mass spectrometry) are accurate but slow, complex, and lack automation for modern supply chains.
• There is a need for high-throughput, intelligent solutions for effective food safety monitoring.

Purpose of the Study:

• To review the application of Multimodal Large Models (MLLMs) in food safety detection.
• To summarize advancements in deep learning, cross-modal intelligence, and multimodal fusion for food safety.
• To discuss current challenges and future directions for intelligent food safety systems.

Main Methods:

• Review of recent advances in Multimodal Large Models (MLLMs).
• Integration of multi-modal perception, knowledge enhancement, and self-supervised pre-training.
• Analysis of progress from deep learning to cross-modal intelligence and fusion mechanisms.

Main Results:

• MLLMs show promise for intelligent food safety by integrating diverse data sources.
• Applications include quality assessment and identification of upstream agricultural risks.
• Progress in multimodal fusion mechanisms enhances detection capabilities.

Conclusions:

• MLLMs present a significant advancement over traditional methods for food safety.
• Future research should focus on data scarcity, reliable intelligence, energy efficiency, and security.
• Intelligent food safety detection systems are crucial for evolving global food systems.