cv

Research Intern, Spatial AI Team

NAVER LABS

Research internship on generalizable 3D Gaussian splatting from unposed sparse-view images.

• Developing a generalizable feed-forward 3D Gaussian splatting model for 3D scene reconstruction from unposed sparse-view (paired-view) images using DUSt3R and MASt3R.

Graduate Researcher

Hanyang SAIL Lab (HY-Vision Lab), Hanyang University

Graduate research on privacy-preserving visual localization and 3D line cloud representations.

• Developed a privacy-preserving 3D line cloud generation method that draws lines by pairing two points, named Paired-Point Lifting (PPL).
• Demonstrated that PPL conceals scene details while maintaining camera pose re-localization accuracy.

Undergraduate Researcher

NIAI Lab, Hanyang University

Deepfake video detection using temporal feature analysis.

• Developed a classification model to discern deepfake videos by training the model to extract abrupt temporal features with a PCA-based loss.

Undergraduate Researcher

ICSP Lab, Hanyang University

Image-based weather classification and restoration.

• Devised a one-for-all model that performs both weather classification and reconstruction under adverse weather conditions such as rain, snow, and haze.

Ph.D

National University of Singapore (NUS)

Computer Science

M.S.

Hanyang University (HYU)

Electronic Engineering

B.S.

Hanyang University (HYU)

Electronic Engineering

Enhancing Sparse-View 3D Reconstruction with Unknown Poses via Geometric Prior Guided Refinement

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshop ULTRRA

Sparse-view 3D reconstruction with unknown camera poses using geometric priors for refinement.

Efficient Privacy-Preserving Visual Localization Using 3D Ray Clouds

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Visual localization framework using 3D ray clouds for efficient privacy preservation.

Paired-Point Lifting for Enhanced Privacy-Preserving Visual Localization

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Privacy-preserving visual localization using paired-point lifting and line cloud representations.

Generalizable 3D Gaussian Splatting with Unposed Images

• Used DUSt3R and MASt3R to estimate geometry and guide 3D Gaussian splatting.
• Focused on generalization across scenes and viewpoints.

Enhancing Privacy-Preserving Visual Localization

• Proposed Paired-Point Lifting (PPL) to generate privacy-preserving line cloud maps.
• Demonstrated robustness against scene recovery attacks while maintaining pose accuracy.

Deep Learning-based Sketched Face Recognition

• Fine-tuned ArcFace for sketched face recognition.
• Augmented sketch data using CycleGAN for improved robustness.

Deepfake Video Detection

• Trained models to capture abrupt temporal features indicative of manipulation.
• Used PCA-based loss to emphasize abnormal variations.

Image-based Weather Classification and Restoration

• Designed a model that handles classification and reconstruction for rain, snow, and haze.

Third Place, IC-PBL+ Graduate Championship 2022-2

IC-PBL Teaching and Learning Center, Hanyang University

Excellence Award, Undergraduate Hanyang Academic

Hanyang University

Excellence Award, The Korean Institute of Broadcast and Media Engineers Fall Conference 2021

The Korean Institute of Broadcast and Media Engineers (KIBME)

HY-in Scholarships (Graduate)

Hanyang University

Graduate scholarship, total amount approximately $12,200.

Dept. of Electronic Engineering Scholarship (Undergraduate)

Hanyang University

Undergraduate scholarship, total amount approximately $14,500.

Conference Reviewer

IEEE / CVF and related communities

Reviewed papers for major computer vision conferences.

• Paper reviewer for ICIP 2022–2023.
• Paper reviewer for WACV 2023–2024.
• Paper reviewer for CVPR 2024.