MINES Lab (Advisor: Hyungseok Kim)Sejong University
Undergraduate ResearcherDec. 2023 - Jan. 2026
Developed and implemented various deep learning models for computer vision tasks, including Human
Action Recognition (HAR) and Object Detection.
Perceptual AI LAB (Advisor: Chanho Eom)Chung‑ang University
Undergraduate ResearcherFeb. 2026 - Present
Project Experiences
Detection of Abnormal Human Activity RecognitionMar. 2024 - Jan. 2025
Developed a system to identify and flag unusual human behaviors from real-time webcam streams for
automated security monitoring.
Engineered a two-stage pipeline: first, utilized YOLOv11 for efficient human
detection, then processed the resulting bounding boxes with a Two-Stream CNN for
detailed action classification.
Multimodal Wearable Visual Aid for the Visually ImpairedFeb. 2025 - Jan. 2026
Contributed to the development of a real-time navigational assistance system by designing and
training models to detect indoor and outdoor obstacles from stereo camera input.
Fine-tuned an RT-DETR model on a custom dataset of outdoor obstacles, achieving
over 90% mAP50 for 5 classes and preparing the model for future deployment on an NPU.
AI Docent Robot for the Visually Impaired (Capstone Design)Mar. 2026 - Jun. 2026
Developed an AI‑powered museum docent robot for visually impaired visitors by integrating autonomous
navigation with voice‑based interactive guidance.
Implemented ROS2‑based autonomous navigation and an LLM‑powered RAG pipeline for personalized
artwork explanations and real‑time Q&A.
Publications
AI Robotic Infant and Toddlers Alert In-Home SystemICCAS 2025
M Kim, S Lee, T Kim and H Kim (oral presentation)Nov. 2025
Developed a unified vision-based system using a VideoMAE V2 backbone and low-light enhancement
filter to recognize 8 baby behaviors, validating its feasibility for real-time robotic monitoring.
Reviewed one of the paper submissions for ICCAS 2025.
Extracurricular Activities
Sejong AI Circle SAIMar. 2025 - Feb. 2026
Member
Led and mentored a study group of 4 junior members, focusing on foundational deep learning concepts
like CNNs and RNNs.
Korea Electronics Technology Institute (KETI)
Research InternshipJan. 2026 - Feb. 2026
Developed an on‑device VLM service using Phi‑4 on Jetson Orin and Raspberry Pi.
Reproduced and implemented a Visual Grounding research paper on Jetson Orin.
Honors & Awards
2025Bronze Award, Sejong AI Challenge 2025SSeoul, S.Korea
2025Gold Award, Sejong AI Challenge 2025FSeoul, S.Korea