Minhyun Lee

Minhyun Lee

AI Researcher, Samsung Electronics AI Center

About

I am a Staff Engineer at Samsung Electronics AI Center, where I work on semiconductor data intelligence and layout pattern generation. My research focuses on practical machine learning systems that remain reliable under limited supervision, noisy labels, and imperfect industrial data pipelines.

Before joining Samsung Electronics, I completed my M.S. and Ph.D. in Integrated Technology at Yonsei University, advised by Prof. Hyunjung Shim. I also interned at NAVER AI Lab with Song Park, Byeongho Heo, and Dongyoon Han. My earlier research centered on weak supervision, semantic segmentation, localization, and representation learning. More recently, I have been working on semiconductor data modeling, generative methods, and robust preprocessing for industrial AI systems.

My current interests include label-efficient learning, robust industrial vision systems, and semiconductor layout generation.

Publications

(* indicates equal contribution; † indicates corresponding author.)

Blind to Position, Biased in Language: Probing Mid-Layer Representational Bias in Vision-Language Encoders for Zero-Shot Language-Grounded Spatial Understanding

NM An*, I Kang*, M Lee, H Shim†

European Conference on Computer Vision (ECCV), 2026

paper

MaskRIS: Semantic Distortion-aware Data Augmentation for Referring Image Segmentation

M Lee*, S Lee*, S Park, D Han, B Heo†, H Shim†

Transactions on Machine Learning Research (TMLR), 2025

paper code

Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part Segmentation

J Choi*, S Lee*, M Lee, S Lee, H Shim†

Conference on Computer Vision and Pattern Recognition (CVPR), 2025

paper code

Understanding Multi-Granularity for Open-Vocabulary Part Segmentation

J Choi*, S Lee*, S Lee, M Lee, H Shim†

Neural Information Processing Systems (NeurIPS), 2024

paper code

SeiT++: Masked Token Modeling Improves Storage-efficient Training

M Lee*, S Park*, B Heo, D Han, H Shim†

European Conference on Computer Vision (ECCV), 2024

paper code

Saliency as pseudo-pixel supervision for weakly and semi-supervised semantic segmentation

M Lee*, S Lee*, J Lee, H Shim†

IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2023

paper

Hybridmatch: Semi-supervised facial landmark detection via hybrid heatmap representations

S Kang*, M Lee*, M Kim, H Shim†

IEEE Access 11, 26125-26135, 2023

paper

Threshold matters in wsss: Manipulating the activation for the robust and accurate segmentation model against thresholds

M Lee*, D Kim*, H Shim†

Conference on Computer Vision and Pattern Recognition (CVPR), 2022

paper code

Railroad is not a train: Saliency as pseudo-pixel supervision for weakly supervised semantic segmentation

S Lee*, M Lee*, J Lee, H Shim†

Conference on Computer Vision and Pattern Recognition (CVPR), 2021

paper code

Psynet: Self-supervised approach to object localization using point symmetric transformation

K Baek*, M Lee*, H Shim†

AAAI Conference on Artificial Intelligence, 2020

paper code

Rethinking Weakly Supervised Semantic Segmentation Utilizing Both Point and Image Labels

M Lee, H Lee, H Shim†

Available at SSRN 5062471, 0

paper

Patents

Weakly supervised semantic segmentation device and method based on pseudo-masks

H Shim, S Lee, M Lee

US Patent No. 11,798,171, granted October 24, 2023. JP Patent App. 2021-207591. Korean Patent App. 10-2021-0124495.

patent

Awards & Honors

Experience

Minhyun Lee
AI Researcher, Samsung Electronics AI Center
based on Martin Saveski's template