Yeho Gwon

I am an undergraduate student at POSTECH, South Korea, majoring in Computer Science and Engineering. I am also working as a research intern at Computer Vision Lab, POSTECH, under the supervision of Prof. Suha Kwak.

My research interests include computer vision and machine learning. I have worked on reducing the annotation cost for image classification tasks by leveraging active learning and using hand-crafted query strategies. Currently, I am interested in vision foundation models (e.g., SAM) and their robustness and generalization beyond the 2D space.

I am looking for internships and research opportunities in the field of machine learning and computer vision. If you are interested in collaborating with me, please feel free to reach out to me!

Email: yeho.gwon@postech.ac.kr

News

Feb 10, 2025	📜 Our paper on designing a novel query for active learning is now up on arXiv. Come and visit to find out how much cost our framework can save!
Jan 3, 2025	📡 I visited CES 2025 in Las Vegas, NV, USA. It was a great experience connecting with many of you during my three-week stay in the US!
Jun 17, 2024	🇺🇸 I started my Summer sessions at University of California, Berkeley. Can’t wait to meet you!
Feb 18, 2022	📚 I started my Bachelor’s course at POSTECH!

Education

Jun, 2024 - Aug, 2024	University of California, Berkeley, Berkeley, CA, USA Visiting Summer Sessions Student
Feb, 2022 - Aug, 2025	Pohang University of Science and Technology (POSTECH), Pohang, South Korea B.S. in Computer Science and Engineering

Experience

Feb, 2023 - Present	Computer Vision Lab at POSTECH, Pohang, South Korea Research Intern Advisor: Prof. Suha Kwak Participate in research projects on active learning for visual recognition tasks.

Publications

* indicates equal contribution.

Enhancing Cost Efficiency in Active Learning with Candidate Set Query

Yeho Gwon*, Sehyun Hwang*, Hoyoung Kim, Jungseul Ok, and Suha Kwak

arXiv preprint, 2025

Abstract Paper Code Project Page

This paper introduces a cost-efficient active learning (AL) framework for classification, featuring a novel query design called candidate set query. Unlike traditional AL queries requiring the oracle to examine all possible classes, our method narrows down the set of candidate classes likely to include the ground-truth class, significantly reducing the search space and labeling cost. Moreover, we leverage conformal prediction to dynamically generate small yet reliable candidate sets, adapting to model enhancement over successive AL rounds. To this end, we introduce an acquisition function designed to prioritize data points that offer high information gain at lower cost. Empirical evaluations on CIFAR-10, CIFAR-100, and ImageNet64x64 demonstrate the effectiveness and scalability of our framework. Notably, it reduces labeling cost by 42% on ImageNet64x64.