- Gyuseong Lee
- M.S. Student
- Website
- Github
- Google Scholar
About
M.S. student in Computer Vision Laboratory at Korea University., Interested in computer vision, especially in diffusion generative models and its controllable generation.
Education
- M.S. in Computer ScienceKorea University, Seoul, Korea2022 - 2024
- B.S. in Electrical EngineeringKorea University, Seoul, Korea2016 - 2022
Publications
- Improving Sample Quality of Diffusion Models Using Self-Attention GuidanceSusung Hong, Gyuseong Lee, Wooseok Jang, Seungryong KimInternational Conference on Computer Vision (ICCV)Oct. 2023
- MIDMs: Matching Interleaved Diffusion Models for Exemplar-based Image TranslationJunyoung Seo*, Gyuseong Lee*, Seokju Cho, Jiyoung Lee, Seungryong KimAAAI Conference on Artificial Intelligence (AAAI)Feb. 2023
- Semi-Supervised Learning of Semantic Correspondence With Pseudo-LabelsJiwon Kim*, Kwangrok Ryoo*, Junyoung Seo*, Gyuseong Lee*, Daehwan Kim, Hansang Cho, Seungryong KimIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)Jun. 2022
- Conmatch: Semi-supervised learning with confidence-guided consistency regularizationJiwon Kim, Youngjo Min, Daehwan Kim, Gyuseong Lee, Junyoung Seo, Kwangrok Ryoo, Seungryong KimEuropean Conference on Computer Vision (ECCV)Oct. 2022
- DiffMatch: Diffusion Model for Dense MatchingJisu Nam, Gyuseong Lee, Sunwoo Kim, Hyeonsu Kim, Hyoungwon Cho, Seyeon Kim, Seungryong KimarXiv preprintMay 2023
- Domain Generalization Using Large Pretrained Models with Mixture-of-AdaptersGyuseong Lee*, Wooseok Jang*, Jin Hyeon Kim, Jaewoo Jung, Seungryong KimarXiv preprintOct. 2023
- Depth-Aware Guidance with Self-Estimated Depth Representations of Diffusion ModelsGyeongnyeon Kim*, Wooseok Jang*, Gyuseong Lee*, Susung Hong, Junyoung Seo, Seungryong KimarXiv preprintDec. 2022
- Towards Flexible Inductive Bias via Progressive Reparameterization SchedulingYunsung Lee*, Gyuseong Lee*, Kwangrok Ryoo*, Hyojun Go*, Jihye Park*, Seungryong KimECCV Workshops on Visual Inductive Priors (VIPriors)Dec. 2022
- AggMatch: Aggregating Pseudo Labels for Semi-Supervised LearningJiwon Kim, Kwangrok Ryoo, Gyuseong Lee, Seokju Cho, Junyoung Seo, Daehwan Kim, Hansang Cho, Seungryong KimarXiv preprintJan. 2022
Skills
- Python3PyTorchGitDockerHTMLJavaScript
Awards & Scholarships
- 2023Best Paper Award, 33th Artificial Intelligence and Signal Processing, IEIESeoul, S.Korea
- 2022Best Paper Award, 32nd Joint Conference of Signal Processing, IEIESeoul, S.Korea
- 2020National Science & Technology Scholarship, Korea Student Aid FoundationSeoul, S.Korea
Projects
- Development of multi-domain generalizable anomaly classification modelSAMSUNG ELECTRO-MECHANICSDOMAIN GENERALIZATION & LABEL EFFICIENT DOMAIN ADAPTATION FOR ANOMALY CLASSIFICATION.Sep. 2022 - Apr. 2023
- Analyzing the effect of model bias on out-of-distribution performance.
- Development of domain generalization algorithm by utilizing knowledge from a largely pre-trained model.
- Efficiently fine-tuning large models to improve out-of-distribution generalization performance.
- Semi-supervised & continual learning for anomaly classificationSAMSUNG ELECTRO-MECHANICSSEMI-SUPERVISED & CONTINUAL LEARNING FOR ANOMALY CLASSIFICATION.Sep. 2022 - Apr. 2023
- Fine-grained, long-tailed semi-supervised learning.
- Propose a contrastive learning based loss function to improve the classification accuracy of defective samples.
- Using a knowledge distillation to maintain performance over time in continual learning settings.
Experiences
- Reviewer of CVPR 2023 and ICCV 20232022 - 2022