Yudam Shin
Research
I study how AI can understand human emotion and behavior, and how these capabilities become responsive interactive systems.
My work spans multimodal emotion recognition, synthetic data for avatar behavior, and real-time interaction in XR. Undergraduate research at Graphics Realization Lab, Chung-Ang University.
Publications
Single-Image-Driven Automated 3D Avatar Generation and Real-Time Motion Retargeting Pipeline for Immersive Exhibition Interaction
전시 환경에서의 단일 이미지 기반 3D 아바타 생성 및 실시간 모션 리타겟팅 파이프라인 연구
Enhancing Deep Learning for Pine Wilt Disease Detection Using 3D Virtual Synthetic Data
SNS 영상 데이터 분석을 통한 전시 후기 분석과 경험 시각화 모델 연구: 릴스(Reels)를 중심으로
불완전한 데이터 상황에서 각성도 및 긍·부정도를 적용한 멀티모달 감정 인식 모델
A Multi-Modal Emotion Recognition Model Incorporating Arousal and Valence with Incomplete Data
각성도 및 긍·부정도를 적용한 멀티모달 감정인식 모델
Adapting the Valence_Arousal Model for Multimodal Emotion Recognition with Incomplete Data
메타버스 내 어뷰징 행동 탐지를 위한 데이터셋 구축 연구
Research & projects
Beyond the Bloom · 병풍연화
An interactive XR exhibition where body movements shape a folktale. Led action recognition, motion retargeting, and Unity system integration within a five-person team.
Read case study ↗ · Digital exhibition ↗TERARiUM
Connected visitor sketches to responsive 3D characters. Built Kinect-to-Unity retargeting, UniRig joint mapping, and face-based visitor re-identification; integrated the team’s generation pipeline.
Read case study ↗Multimodal emotion recognition
Studied arousal and valence representations for speech–text emotion recognition with incomplete modalities. Led model design, implementation, experiments, and analysis.
Read case study ↗Metaverse behavior & synthetic data
Contributed to defining abusive avatar behaviors, a participant survey, and synthetic-data production using motion capture, motion generation, Blender, and Unity Perception.
Read case study ↗Mindy
Built an MVP for child–parent emotional communication, from parent interviews and product research to a Flutter application with voice conversation and parent reports.
Read case study ↗Echotion
Connected speech recognition, KoBERT emotion classification, and expressive subtitles in a desktop interface.
Read case study ↗