Minhyun Lee

Publications

* denotes equal contribution · † denotes corresponding author

Conference Proceedings

Qualitative comparison of text-to-image safety methods, with sensitive regions masked and CALM outputs in the rightmost column.

Keep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation

NaHyeon Park, Minhyun Lee, Hyunjung Shim†
Advances in Neural Information Processing Systems
NeurIPS 2026
PaperCode
Vision and language encoder probing, layer selection, and spatial grounding.

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

Na Min An*, Inha Kang*, Minhyun Lee, Hyunjung Shim†
Proceedings of the 2026 European Conference on Computer Vision
ECCV 2026
PaperCode
Source images, object and part correspondence maps, and part segmentation predictions.

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

Jiho Choi*, Seonho Lee*, Minhyun Lee, Seungho Lee, Hyunjung Shim†
Proceedings of the 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition
CVPR 2025
PaperCode
A dog segmented into object-level and part-level regions, with part attention maps.

Understanding Multi-Granularity for Open-Vocabulary Part Segmentation

Jiho Choi*, Seonho Lee*, Seungho Lee, Minhyun Lee, Hyunjung Shim†
Advances in Neural Information Processing Systems
NeurIPS 2024
PaperCode
Accuracy and training-data storage comparison from the SeiT++ paper.

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

Minhyun Lee*, Song Park*, Byeongho Heo, Dongyoon Han, Hyunjung Shim†
Proceedings of the 2024 European Conference on Computer Vision
ECCV 2024
PaperCode
Ground-truth masks and class activation maps before and after activation manipulation.

Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds

Minhyun Lee*, Dongseob Kim*, Hyunjung Shim†
Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition
CVPR 2022
PaperCode
Image labels and a saliency map provide complementary signals for a segmentation mask.

Railroad Is Not a Train: Saliency As Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation

Seungho Lee*, Minhyun Lee*, Jongwuk Lee, Hyunjung Shim†
Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition
CVPR 2021
PaperCode
Bird images with predicted localization boxes and activation maps.

PsyNet: Self-Supervised Approach to Object Localization Using Point Symmetric Transformation

Kyungjune Baek*, Minhyun Lee*, Hyunjung Shim†
Proceedings of the 2020 AAAI Conference on Artificial Intelligence
AAAI 2020
PaperCode

Journal Articles

Comparison of augmentation effects on semantic segmentation and language-guided segmentation.

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

Minhyun Lee*, Seungho Lee*, Song Park, Dongyoon Han, Byeongho Heo†, Hyunjung Shim†
Transactions on Machine Learning Research
TMLR 2025
PaperCode
EPS++ framework combining localization maps, saliency supervision, refinement, and image-level classification.

Saliency as Pseudo-Pixel Supervision for Weakly and Semi-Supervised Semantic Segmentation

Minhyun Lee*, Seungho Lee*, Jongwuk Lee, Hyunjung Shim†
IEEE Transactions on Pattern Analysis and Machine Intelligence
IEEE TPAMI 2023
Paper
HybridMatch framework combining 2-D and 1-D heatmap predictions for facial landmarks.

HybridMatch: Semi-Supervised Facial Landmark Detection via Hybrid Heatmap Representations

Seoungyoon Kang*, Minhyun Lee*, Minjae Kim, Hyunjung Shim†
IEEE Access, Volume 11
IEEE Access 2023
Paper

Preprints

NCLS architecture combining image-level pseudo masks and point labels with consistency-based noise filtering; original preprint watermark retained.

Rethinking Weakly Supervised Semantic Segmentation Utilizing Both Point and Image Labels

Minhyun Lee, Hwijeong Lee, Hyunjung Shim†
Preprint available on SSRN
SSRN preprint
Paper

Patents

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

Hyunjung Shim, Seungho Lee, Minhyun Lee
US Patent No. 11,798,171, granted October 24, 2023. JP Patent App. 2021-207591. Korean Patent App. 10-2021-0124495.
US Patent 11,798,171 · 2023
Patent