Machine Learning Researcher in Generative Modeling and Computer Vision
I am a machine learning researcher specialized in generative models (diffusion models, GANs), synthetic data generation, and multimodal deep learning. I received my PhD (Dr. rer. nat.) in Computer Science Magna cum Laude from Heidelberg University under the supervision of Prof. Dr. Carsten Rother.
Currently, I lead R&D group at AIMMO Germany GmbH, where I headed three teams focused on synthetic data pipelines, auto-labeling systems, and international research collaborations. Previously, I conducted research at Robert Bosch GmbH on 3D mesh generation, neural rendering, and lighting estimation for automotive applications.
Jun 2024 — Graduated with Dr. rer. nat. in Computer Science (Magna cum Laude), Heidelberg University.
2024 — Published Image Generation of Ultra-Thin Polymer Films Using Diffusion Models at ICEIC 2024.
2024 — Published Fault Diagnosis of Indium Tin Oxide Electrodes by Multi-channel S-parameter Patterns at ICEIC 2024.
2023 — Appointed Head of R&D Group at AIMMO Germany GmbH.
2022 — Published Spatio-Temporal Outdoor Lighting Aggregation on Image Sequences using Transformer Networks in IJCV.
2021 — Presented Spatiotemporal Outdoor Lighting Aggregation on Image Sequences at GCPR 2021.
2021 — Published CEGAN: Classification Enhancement GANs in Neural Networks.
2019 — Published Generative Oversampling Method for Imbalanced Data on Bearing Fault Detection and Diagnosis in Applied Sciences.
2019 — Joined Robert Bosch GmbH, Hildesheim as a Doctoral Researcher.
2019 — Published Robust Shipping Label Recognition for Logistics at ICIP 2019.
2018 — Presented Real-time Apparent Resolution Enhancement for HMDs at ACM SIGGRAPH i3D 2018.
2018–2020 — Awarded LG Electronics Scholarship.
Tae Yeob Kang, Haebom Lee, Sungho Suh
IEEE Transactions on Device and Materials Reliability 2024
Tae Yeob Kang, Haebom Lee, Sungho Suh
IEEE Transactions on Device and Materials Reliability 2024
Sungho Suh, Haebom Lee, Tae Yeob Kang
International Conference on Electronics, Information, and Communication (ICEIC) 2024
Sungho Suh, Haebom Lee, Tae Yeob Kang
International Conference on Electronics, Information, and Communication (ICEIC) 2024
Haebom Lee, Christian Homeyer, Robert Herzog, Jan Rexilius, Carsten Rother
International Journal of Computer Vision 2022
Haebom Lee, Robert Herzog, Jan Rexilius, Carsten Rother
DAGM German Conference on Pattern Recognition 2021
Sungho Suh, Haebom Lee, Paul Lukowicz, Yong Oh Lee
Neural Networks 2021
Sungho Suh, Haebom Lee, Yong Oh Lee, Paul Lukowicz, Jongwoon Hwang
IEEE International Conference on Image Processing 2019
Sungho Suh*, Haebom Lee*, Jun Jo, Paul Lukowicz and Yong Oh Lee
Applied Sciences 2019, 9(4), 746Â
Haebom Lee, Jun Jo, Yong Oh Lee, Nuriye Korkmaz Zirpel, Leon Abelmann
International Conference on Practical Applications of Computational Biology & Bioinformatics. Springer, Cham, 2018.Â
Presented at International Conference on PACBB 2018, Toledo, Spain
Haebom Lee, Piotr Didyk
Proceedings of the ACM on Computer Graphics and Interactive Techniques 1.1 (2018): 19.
Presented at ACM SIGGRAPH Symposium on Interactive 3D Graphics and Games 2018, Montreal, Canada
Giljoo Nam, Haebom Lee, Sungsoo Oh, Min H. Kim
IEEE Transactions on Instrumentation and Measurement, Oct. 19, 2015, 65(2), pp.297-304
Haebom Lee, Min H. Kim
Proc. Int. Conf. Image and Signal Processing 2014, Lecture Notes in Computer Science, 8509, June 2014, pp. 26-34 (oral presentation)
Byungsoo Kim, Haebom Lee, Kee-Eung Kim
Undergraduate Research Participation (URP) Program, KAIST, 2008 Fall
Kaggle is well known for hosting deep learning related competitions. Iceberg Classifier Challenge was one of them, encouraging researchers to distinguish icebergs from ships by examining a set of radar images. I designed a CNN model as well as image processing algorithms for preprocessing.
One of the most important techniques for autonomous driving is image segmentation. Exploiting deep learning in this area is already common because of its accuracy and flexibility. In this project, I tested several state-of-the-art deep learning segmentation algorithms. Furthermore, an efficient method to generate street view images from a simulated scene was also investigated.
There have been several apparent resolution enhancement techniques for high-framerate displays using an expensive optimization or a vibrating motor. Those characteristics made the methods unable to be performed on head mounted displays, which are usually showing real-time contents. The proposed technique approximates the time consuming optimization using a set of filters. While there is a slight loss of clarity, it improves apparent resolution in real time as shown in a user experiment. The details will be provided in this page.
In order to make the smartphone display inspection more reliable, quantified, and automatized, a system based on HDR imaging and image appearance modeling was developed for LG Electronics. The system efficiently classified faulty products, which are containing yellowish regions or black dots. The details are presented in this paper.
PACT Engine is an OpenGL-based 3D game engine developed in Com2uS, a mobile game company. The engine consists of a viewer, a 3ds Max plugin and libraries for iOS and Android. It supports keyframe-based animations for bones, vertices and textures
An OS-independent development environment gets rid of problems occur in porting a game from one platform to another. The 'Wrapper' project was therefore initiated in Com2uS to enable the developers to focus only on a game itself. The project is categorized into 'iOS Wrapper' and 'Android Wrapper'. I took the charge of the Android wrapper and successfully launched the first Android game of Com2uS.
I participated as an engine programmer in several game developments: Homerun Battle 3D, Homerun Battle 2, Heavy Gunner 3D, Zombie Runaway, Swing Shot, and Slice It!.
Netflix set $1,000,000 prize for the accurate recommendation system. I implemented k-Nearest Neighbors, k-Means Clustering, and Naive Bayesian Classifier on a Hadoop-based distributed system to participate in the Netflix Prize.