Jin-Seop Lee

I am a Ph.D. candidate in the Information & Intelligence System Lab (IISLab) at Sungkyunkwan University, supervised by Prof. Jee-Hyong Lee. I received my Bachelor's degree from Sungkyunkwan University.

Email / CV / Google Scholar / LinkedIn / Github

News

[2025-08] I received a Ph.D fellowship from NRF(National Research Foundation) of Korea.
[2025-07] Two papers accepted to BMVC 2025.
[2025-05] One paper accepted to ACL 2025.
[2024-12] One paper accepted to AAAI 2025.
[2024-10] One paper accepted to WACV 2025.
[2024-07] Two papers accepted to ECCV 2024.
[2024-02] One paper accepted to CVPR 2024.
[2024-01] One paper accepted to EAAI 2024.

Research

My research interests include various tasks in computer vision, deep learning, and machine learning. I have been conducted research on data-efficient learning methods, including self-/semi-supervised learning, weakly-supervised learning, and domain generalization, etc. I am currently conducting research to improve the understanding and reasoning abilities of multi-modal large language models (MLLMs). Also, I am interested in multi-modal learning, video understanding, image/video synthesis, efficient AI systems, and streaming video LLMs.

	CountCluster: Training-Free Object Quantity Guidance with Cross-Attention Map Clustering for Text-to-Image Generation Joohyeon Lee, Jin-Seop Lee, Jee-Hyong Lee Preprint [Project page] [Code] [arXiv]
	TAG: A Simple Yet Effective Temporal-Aware Approach for Zero-Shot Video Temporal Grounding Jin-Seop Lee, Sungjoon Lee, Jaehan Ahn, Yunseok Choi, Jee-Hyong Lee BMVC, 2025 [Project page] [Code] [arXiv] TAG effectively captures the temporal context of videos and addresses distorted similarity distributions without training.
	Stabilizing Open-Set Test-Time Adaptation via Primary-Auxiliary Filtering and Knowledge-Integrated Prediction Byung-Joon Lee, Jin-Seop Lee, Jee-Hyong Lee BMVC, 2025 [Project page] [Code] [arXiv] OSTTA employs an auxiliary filter to validate data and calibrates the outputs of the adapting model, EMA model, and source model to integrate their complementary knowledge.
	DCG-SQL: Enhancing In-Context Learning for Text-to-SQL with Deep Contextual Schema Link Graph Jihyung Lee, Jin-Seop Lee, Jaehoon Lee, YunSeok Choi, Jee-Hyong Lee ACL, 2025 [Project page] [Code] [arXiv] DCG-SQL improves text-to-SQL generation by incorporating a deep contextual schema link graph that captures key elements and semantic relationships between the question and database schema.
	DomCLP: Domain-wise Contrastive Learning with Prototype Mixup for Unsupervised Domain Generalization Jin-Seop Lee, Noo-ri Kim, Jee-Hyong Lee AAAI, 2025 [Project page] [paper] [Code] [arXiv] DomCLP enhances generalization to unseen domains by combining domain-aware contrastive learning with prototype mixup to learn robust and domain-invariant features.
	Feature-level and Spatial-level Activation Expansion for Weakly-Supervised Semantic Segmentation Junsu Choi, Jin-Seop Lee, Noo-ri Kim, SuHyun Yoon, Jee-Hyong Lee WACV, 2025 [Paper] [Code] FSAE improves weakly-supervised segmentation by expanding Class Activation Maps along feature and spatial dimensions to better capture full object regions.
	IGNORE: Information Gap-based False Negative Loss Rejection for Single Positive Multi-Label Learning Gyeong Ryeol Song, Noo-ri Kim, Jin-Seop Lee, Jee-Hyong Lee ECCV, 2024 [Paper] [Code] IGNORE identifies and filters hidden positive labels using information gaps from pseudo masks to reduce false negatives in single-positive multi-label learning.
	ExMatch: Semi-Supervised Learning with Scarce Labeled Samples with Additional Exploitation of Unlabeled Samples Noo-ri Kim, Jin-Seop Lee, Jee-Hyong Lee ECCV, 2024 [Paper] ExMatch boosts semi-supervised learning with extremely limited labels by selectively leveraging confident unlabeled samples for self-training.
	Learning with Structural Labels in Learning with Noisy Labels Noo-ri Kim, Jin-Seop Lee, Jee-Hyong Lee CVPR, 2024 [Paper] LSL improves noisy label learning by incorporating structural information from data distribution to prevent overfitting to incorrect labels.
	Automation of Trimming Die Design Inspection by Zigzag Process Between AI and CAD Domains Jin-Seop Lee, Tae-Hyun Kim, Sang-Hwan Jeon, Sung-Hyun Park, Sang-Hi Kim, Eun-Ho Lee, Jee-Hyong Lee EAAI(Engineering Applications of Artificial Intelligence), 2024, TOP 3%, IF 8.0 [Paper] Our zigzag process automates trimming die inspection through alternating AI and CAD collaboration, achieving high accuracy and significantly reduced inspection time.
	Automatic defect classification using semi-supervised learning with defect localization Yusung Kim, Jin-Seop Lee, Jee-Hyong Lee IEEE Transactions on Semiconductor Manufacturing, 2023 [Paper] Our method achieves robust defect classification in semiconductor manufacturing by combining localization-guided detection with semi-supervised learning.