Jo Hyeongseob
Logo M.S. student at Dongguk University

I am from South Korea. I am currently a Master’s student in the Department of Computer Science and Engineering at Dongguk University. I am a member of the Smart Vision and Media (SVM) Lab, led by Professor Sung In Cho. I received my Bachelor of Science in Mechanical Engineering from Dongguk University (2024).

My research interests include computer vision, image processing, defect classification, and anomaly detection, with a focus on solving real-world problems. In particular, I aim to develop computer vision applications for industrial domains to enhance product quality in automated manufacturing environments.


News
2025
Submitted a paper on defect classification to a conference Under Review
May 17
Submitted a paper on unsupervised binary hashing to a journal Under Review
May 03
2024
Our paper "A Survey of Unsupervised Learning-Based Out-of-Distribution Detection" has been published in ICCE-Asia. Published
Nov 03
Started industry-academic collaboration with LG Display on defect inspection Read more
Mar 01
2023
Jul 01
Selected Publications (view all )
A Survey of Unsupervised Learning-Based Out-of-Distribution Detection
A Survey of Unsupervised Learning-Based Out-of-Distribution Detection

Hyeongseob Jo, Seunggi Park, Sung In Cho

IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) 2024

Out-of-distribution (OOD) detection is the task of distinguishing abnormal data that lies outside the bounds of the training dataset’s distribution. OOD detection plays a vital role in various applications of machine learning and deep learning, including intrusion detection in cybersecurity, diagnostics in medical data, and defect classification in manufacturing processes. While models for OOD detection are typically trained using supervised learning, this approach requires significant cost and effort such as collection and labeling of OOD data. To address this issue, unsupervised learning-based methods have been proposed, which can overcome the drawbacks of supervised approaches. In this paper, we introduce generative model-based OOD methods and self-supervised OOD detection methods within the realm of unsupervised learning.

A Survey of Unsupervised Learning-Based Out-of-Distribution Detection

Hyeongseob Jo, Seunggi Park, Sung In Cho

IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) 2024

Out-of-distribution (OOD) detection is the task of distinguishing abnormal data that lies outside the bounds of the training dataset’s distribution. OOD detection plays a vital role in various applications of machine learning and deep learning, including intrusion detection in cybersecurity, diagnostics in medical data, and defect classification in manufacturing processes. While models for OOD detection are typically trained using supervised learning, this approach requires significant cost and effort such as collection and labeling of OOD data. To address this issue, unsupervised learning-based methods have been proposed, which can overcome the drawbacks of supervised approaches. In this paper, we introduce generative model-based OOD methods and self-supervised OOD detection methods within the realm of unsupervised learning.

All publications
Completed Projects (view all )
Development of a Multi-illumination Visual Inspection Algorithm
Development of a Multi-illumination Visual Inspection Algorithm

LG Display x SVM Lab [Industry-Sponsored Research]

2024.03–2025.02

✅ Defect Classification Accuracy: 98.5% (192/195 images correctly classified on test set).
✅ Normal Classification Accuracy: 51.3% (157/306 images correctly classified on test set).
✅ Deployment: The final algorithm developed in this project has been deployed in a real-world LG Display inspection system.
✅ Paper in Progress: Paper has been submitted and is under review (BMVC 2025).

Development of a Multi-illumination Visual Inspection Algorithm

LG Display x SVM Lab [Industry-Sponsored Research]

2024.03–2025.02

✅ Defect Classification Accuracy: 98.5% (192/195 images correctly classified on test set).
✅ Normal Classification Accuracy: 51.3% (157/306 images correctly classified on test set).
✅ Deployment: The final algorithm developed in this project has been deployed in a real-world LG Display inspection system.
✅ Paper in Progress: Paper has been submitted and is under review (BMVC 2025).

All projects