Projects

Keyword : Multimodal Deep Learning, Sensor Fusion, In Cabin Sensing

In collaboration with IITP and VAIV Company, our lab is developing a VAIVGeM-based Multi-Modal LLM framework designed for real-time disaster response in smart cities. By integrating heterogeneous data from IoT sensors, vision, and text with RAG (Retrieval-Augmented Generation) technology, we effectively minimize hallucinations and maximize information reliability. Specifically, we employ a hierarchical Multi-Agent system to analyze diverse data streams—including SNS and news—enabling the provision of optimized evacuation routes and tailored safety guidelines for individual users. This project aims to establish an intelligent safety system capable of real-time inference on edge devices, ensuring swift action and securing golden time during emergencies. 

Keyword : Smart City, Disaster Detection, Multimodal LLM, RAG, Edge AI, Multi Agent System

In collaboration with the National Institute of Animal Science and industry partners, our research focuses on advancing a data-driven traceability and quality management framework for Hanwoo beef. Our work integrates heterogeneous data sources—including on-farm environmental sensors, feeding behavior records, growth trajectories, and slaughterhouse carcass data—into a unified analytics platform to enable precise prediction of meat quality indicators such as shear force (WBSF), backfat thickness, feeding loss, and carcass yield. By leveraging machine learning and deep learning models tailored to livestock data characteristics, our team develops cut-specific tenderness prediction models, backfat estimation algorithms, and behavior-aware productivity analysis pipelines. Furthermore, production-to-slaughter linkage analysis is employed to quantify how environmental stressors (e.g., THI), feeding patterns, and growth dynamics influence carcass outcomes and grading distributions. This project aims to establish a scalable, explainable, and field-validated Hanwoo traceability system that supports intelligent decision-making for producers, enhances meat quality consistency, and contributes to the digital transformation of smart livestock management.

Keyword: Smart Livestock, Meat Quality Prediction