Trustworthy & Uncertainty-Aware AI
Building models with calibrated confidence, interpretable uncertainty, and greater robustness under distribution shift and adversarial inputs.
AI Researcher · Myongji University
Trustworthy AI for open-world and physical systems.
I’m Yongho Kim 김용호, an AI researcher at DAN Lab, Myongji University. My research spans uncertainty-aware learning, open-world visual recognition, efficient deployment, and emerging embodied AI.
Currently
Ph.D./M.S. Combined
DAN Lab · 2021.03—
Current focusTrustworthy Visual AI
Based atMyongji University
Published26 Papers
Invented7 Patent applications
Profile
Developing reliable visual intelligence for real-world environments.
My research connects computer vision, uncertainty, and efficient computing. I enjoy turning careful experiments into models that are easier to trust—and practical enough to move beyond controlled benchmarks. I am also exploring how reliable learning can transfer safely across heterogeneous embodied systems.
Research directions
Current research interests
My work addresses reliable prediction, recognition of unfamiliar objects, efficient inference, and safe knowledge transfer into physical systems.
Open-World Visual Recognition
Helping visual models recognize known and unseen objects through open-vocabulary learning, OOD detection, and anomaly detection.
Efficient Edge Vision
Optimizing real-time visual intelligence for resource-constrained accelerators and practical deployment environments.
Physical AI & Cross-Embodiment Learning
An emerging research direction exploring safe knowledge transfer across heterogeneous robots and embodied systems.
Funded research
Selected funded research
Research projects connected to my co-authored publications and co-invented patents.
AI-MAC Protocol on Distributed Machine Learning for Intelligent Flying Base Station
- Agency
- NRF
- Project ID
- 2022R1A2C2005705
XAI · Federated learning · UAV / AI-MAC
Linked through publications & patentsAdvanced Core Technologies for WLAN Based on eXplainable AI
- Agency
- IITP · MSIT
- Project ID
- 2021-0-00990
XAI · Weakly supervised localization · Video analysis
Linked research outputs6G STAR-MAC
- Agency
- IITP · MSIT
- Project ID
- 2021-0-00368
Federated learning · Random access · Autonomous MAC
Linked through publications & patentsVision Recognition SDK Based on Domestic AI Semiconductors
- Agency
- KEIT · MOTIE
- Project ID
- RS-2024-00469138
Point cloud · OOD / uncertainty · Vision SDK
Linked research outputsAutonomous VR/AR Content Generation Reflecting Usage Environment
- Agency
- IITP · MSIT
- Project ID
- 2020-0-00994-003
Indoor positioning · Wireless + image fusion
Linked through publications & patentsMassive IoT Security from Cloud to Edge Computing
- Agency
- NRF
- Project ID
- 2019R1F1A1060742
Edge AI · Federated learning · Video analysis
Linked research outputsProject context cross-checked with DAN Lab research and connected publication records.
Intellectual property
Patent portfolio
07
Original Korean applications
Applicant · Myongji University Industry and Academia Cooperation Foundation
-
2025Application filed
미지의 객체를 자동으로 레이블링하는 시스템 및 방법
Application 10-2025-0024785 · Filed 2025-02-26
Inventors · 박현희, 김용호
Applicant · 명지대학교 산학협력단
DAN Lab publication archive -
2024Pending
속성 마스크 압축 기술을 이용한 설명 맵 생성 시스템 및 방법
Application 10-2024-0027911 · Publication KR20250131432A · Filed 2024-02-27
Inventors · 박현희, 김용호
Applicant · 명지대학교 산학협력단
Public patent record -
2022Registered
로컬 모델 최적화 기반의 연합학습을 위한 시스템, 이를 위한 장치 및 이를 위한 방법
Application 10-2022-0190794 · Publication KR20240107807A · Registration 10-2964363
Filed 2022-12-30 · Registered 2026-05-08
Inventors · 박현희, 유철우, 김형빈, 김용호
Applicant · 명지대학교 산학협력단
-
2022Registered
연합학습 프레임워크에서 효율적 에너지 소비 및 정확도 향상을 위한 모델을 학습하기 위한 장치 및 이를 위한 방법
Application 10-2022-0190795 · Publication KR20240107808A · Registration 10-2986185
Filed 2022-12-30 · Registered 2026-06-30
Inventors · 박현희, 유철우, 김형빈, 김용호
Applicant · 명지대학교 산학협력단
-
2022Registered
무선 신호와 이미지를 이용하는 실내 측위 시스템 및 이를 수행하는 방법
Application 10-2022-0162821 · Publication KR20240079680A · Registration 10-2981004
Filed 2022-11-29 · Registered 2026-06-22
Inventors · 유철우 외 3인 (김용호 포함)
Applicant · 명지대학교 산학협력단
-
2022Application ceased
데이터 레이블링 시스템 및 방법
Application 10-2022-0024962 · Publication KR20230127538A · Filed 2022-02-25
Inventors · 박현희, 김용호
Applicant · 명지대학교 산학협력단
Public patent record -
2021Pending divisional
딥러닝을 이용한 스포츠 경기 동영상의 상황 변화 구간 추출 방법 및 장치
Original application 10-2021-0191387 · Publication KR20230101353A · Filed 2021-12-29
Divisional application 10-2024-0095448 · Publication KR20240117511A · Filed 2024-07-19
Inventors · 박현희, 김형빈, 우기문, 김용호, 김지하
Applicant · 명지대학교 산학협력단
Public patent record
Patent status cross-checked with DAN Lab publication archive , public patent records, and official university registration records.
Research record · 2021—2026
Publications
26journal & conference papers
Showing all 26 publications
- 2026Conference
소수 샘플 기반 이상 탐지에서 학습 기준 클래스에 따른 탐지 성능 비교
36회 JCCI
- 2026Conference
피셔 정보량 기반의 적응형 정규화를 통한 증거 기반 딥러닝의 불확실성 추정 성능 향상
한국통신학회 동계종합학술발표대회
- 2025Journal
FedWT: Federated Learning with Minimum Spanning Tree-based Weighted Tree Aggregation for UAV Networks
ICT Express, vol. 11, no. 2, pp. 275–280
- 2025Conference
Adaptive ROI-Aware Point Cloud Downsampling for 3D Media Transmission
International Conference on BWCCA
- 2025Conference
Query Embedding Visualization in a PROB-Based Framework for Open-World Object Detection
International Conference on IMIS
- 2025Conference
Open-World Object Detection을 위한 확률적 객체성 기반 Unknown 객체 자동 레이블링 기법 제안
한국통신학회 동계종합학술발표대회
- 2024Journal
Integrated Indoor Positioning Methods to Optimize Computations and Prediction Accuracy Enhancement
Computational Intelligence, vol. 40, no. 1
- 2024Journal
Dynamic Transmission and Delay Optimization Random Access for Reduced Power Consumption
IEEE Access, vol. 12, pp. 55033–55050
- 2024Conference
컴퓨터 비전을 위한 설명가능한 인공지능 연구동향
34회 JCCI
- 2023Journal
K-FL: Kalman Filter-based Clustering Federated Learning Method
IEEE Access, vol. 11, pp. 36097–36105
- 2023Conference
Enhancing Image Classification and Explainability with Object Isolation and Background Randomization
International Conference on 3PGCIC
- 2023Conference
SISE의 연산량 감소를 위한 다양한 attribution mask compress 기법 제안 및 성능 비교
33회 JCCI
- 2023Conference
5G mMTC 환경에서 2-pair UE NOMA를 이용한 sectorization 랜덤 액세스 절차의 성능 비교
한국통신학회 동계종합학술발표대회
- 2022Journal
대규모 연합학습을 위한 효율적 분산 클러스터링 기법
한국통신학회논문지, vol. 47, no. 1, pp. 198–205
- 2022Journal
Light-CAM: 임베디드 단말의 Weakly Supervised Object Localization을 위한 경량화 모델
한국통신학회논문지, vol. 47, no. 8, pp. 1144–1152
- 2022Conference
Hierarchical Output Model of CNN Learning using Multi Label Datasets
International Conference on IMIS
- 2022Conference
다중 클래스 데이터 셋에서의 딥러닝 성능 개선을 위한 다중 출력 모델 제안
32회 JCCI
- 2022Conference
CAM의 IoU 정확도 개선을 위한 재학습 기반 Discriminative Localization
한국통신학회 동계종합학술발표대회
- 2021Conference
Reducing Model Cost Based on the Weights of Each Layer for Federated Learning Clustering
International Conference on ICUFN
- 2021Conference
CAM의 바운딩 박스 IoU 개선을 위한 Background Remove-Average Class Activation Map 기법 제안
한국통신학회 추계종합학술발표대회
- 2021Conference
MEC 환경에서 5G-V2X 기반 사고 예방 시스템 제안
한국통신학회 하계종합학술발표대회
- 2021Conference
사용자별 가중치 표준 편차를 활용한 Federated Learning 성능 향상 기법
한국통신학회 하계종합학술발표대회
- 2021Conference
NHANES 데이터를 이용한 생활 패턴 기반의 우울증 예측 머신러닝 기법
한국통신학회 하계종합학술발표대회
- 2021Conference
캡션 검출을 활용한 영상 내 특정 시점 자동 검출 기법
31회 JCCI
- 2021Conference
템플릿 매칭을 이용한 야구 하이라이트 추출 기법
31회 JCCI
- 2021Conference
투구 장면 탐지 모델을 위한 Data Labeling 기법
31회 JCCI
Publication record cross-checked with the DAN Lab publication archive and publisher records.
Contact
Research conversations and collaboration.
I welcome conversations with researchers and practitioners working on reliable, open-world, efficient, or embodied AI.
yhkim98@mju.ac.kr