Yongho Son

I'm a second-year PhD student at Hanyang University, advised by Prof.Je Hyeong Hong. I also completed my Master's degree at Hanyang University under the guidance of Prof.Jun Won Choi.

Email / CV / Github

Research

Hello, I am Yongho Son. I am currently pursuing a Ph.D. in the Artificial Intelligence program at Hanyang University. Prior to that, I received a Bachelor of Science degree in Computer Science from Stony Brook University, graduating with cum laude honors. My primary research interests lie in the domain of computer vision, with an active focus on anomaly detection and object detection.

	Perturb, Aggregate then Reconstruct: Consensus Aggregation and Fourier-Attended Multi-Decoders for Reconstruction-based Multi-View Anomaly Detection Yongho Son, Junyeong Yun, Je Hyeong Hong ACM International Conference on Multimedia , 2026 State-of-the-art multi-view anomaly detection performance in Real-IAD and MANTA dataset. Introduced a feature perturbation strategy coupled with a consensus aggregation module to construct highly aligned pseudo-anomaly representations.
	Long-tailed Detection and Classification of Wafer Defects from Scanning Electron Microscope Images Robust to Diverse Image Backgrounds and Defect Scales Taekyeong Park, Yongho Son, Sanghyuk Moon, Seungju Han, Je Hyeong Hong Engineering Applications of Artificial Intelligence (IF 8.0, Categorical JCR < 2.6%)*, 2025 State-of-the-art defect detection and classification performance in wafer dataset with other public industrial anomaly benchmarks. Our method operates at a speed of approximately 42 fps, which allows for time-efficient inspection.
	FUN-AD: Fully Unsupervised Learning for Anomaly Detection with Noisy Training Data Jiin Im, Yongho Son, Je Hyeong Hong The Winter Conference on Applications of Computer Vision (WACV), 2025 State-of-the-art performance across different industrial anomaly benchmarks in presence of contaminated training data. Our method operates at an impressive speed of approximately 113 fps, outperforming other methods.
	PIDDNet: RGB-Depth Fusion Network for Real-time Semantic Segmentation Yunsik Shin, Yongho Son, Junghee Park, Chaehyun Lee, YangGon Kim, Jun Won Choi International Conference on Information and Communication Technology Convergence (ICTC), 2023 Realtime RGB-Depth sensor fusion based semantic segmentation through bidirectional fusion technique which operates on Jetson Orin at 20 FPS.
	3D Object Detection using Dynamic • Static Motion Information from Multi-camera Image Sequences Yongho Son, Jungho Kim, Jun Won Choi THE 33rd JOINT CONFERENCE ON COMMUNICATIONS AND INFORMATION (JCCI), 2023 Proposed a recurrent neural network that performs motion compensation with a new gate called "motion gate".