Jiin Im

Jiin Im

MS Student @ Spatial AI Lab, Hanyang University

About

I am a Master's student at the Spatial AI Lab, Hanyang University, advised by Prof. Je Hyeong Hong. My research interests lie in semantic correspondence, spatial reasoning, and geometry-aware vision.

I am particularly interested in how intelligent systems can discover structural relationships across different visual observations, and how such correspondence can serve as a foundation for spatial understanding. Rather than relying solely on appearance, I aim to build models that capture meaningful relations across entities through unsupervised or weakly supervised learning.

My goal is to enable machines to reason over spatial structure, infer geometric relationships, and interpret scenes in a way that generalizes across tasks and environments. This line of research is motivated by challenges such as partial observability, viewpoint variation, and the need for functional understanding, which arise frequently in real-world perception and robotics.

Education

Hanyang University Mar. 2024 - Present

M.S. in Electronic Engineering (Advisor: Je Hyeong Hong)

Hanyang University Mar. 2020 - Feb. 2024

B.S. in Electronic Engineering, Summa Cum Laude

Publications

FUN-AD Teaser Image

FUN-AD: Fully Unsupervised Learning for Anomaly Detection with Noisy Training Data

Jiin Im*, Yongho Son*, Je Hyeong Hong

WACV 2025

Projects

Semantic Correspondence via Structural Information on 2D and 3D Domains Mar. 2025 - Present

Establishing semantic correspondence on both 2D image and 3D geometry domains using structural information, which is related to a paper currently under review.

Non-Rigid Registration of Deformed Point Clouds via Geometric Correspondence Jun. 2025 - Dec. 2025

Focused on estimating geometric correspondence in deformed point clouds to perform robust non-rigid registration.

Privacy-Preserving Geometric Description against Image Reconstruction Jun. 2024 - Feb. 2025

Explored geometric description methods that prevent original image recovery from descriptors while maintaining matching performance.

Unsupervised Anomaly Detection with Noisy Training Data Jan. 2023 - May. 2024

Addressing anomaly detection in unsupervised settings with noisy training data related to FUN-AD.

Experience

Hobbies

In my free time, I enjoy:

Contact

βœ‰οΈ Email: skqldit33@hanyang.ac.kr

πŸ“ Location: Engineering Center Annex Unit 415-1, Hanyang University, Seongdong-gu, Wangsimni-ro 222, Seoul, South Korea