Projects

Retrieval-Augmented Dense Video Captioning and QA System

#Dense Video Captioning #Retrieval-Augmented Generation #Video QA

End-to-end pipeline for automated highlight extraction, dense captioning, and natural-language QA over long-form (1hr+) videos. Funded by Piaspace (2025).

Risk State Prediction Model Using Construction Site Images

#Vision-Language Model #Safety #Fine-tuning

VLM-based risk-state prediction system fine-tuned to detect unsafe worker behavior and equipment hazards. Funded by Doosan Enerbility (2025).

Zero-Shot Captioning for Driver Status Reporting Agents

#Zero-shot Captioning #Driver Monitoring #Vision-Language

Zero-shot image captioning for in-vehicle driver monitoring, generating natural-language descriptions of driver states. Funded by Hyundai NGV (2024).