PulsePath: Two-Stage Generative Waveform Refinement for Robust Remote Photoplethysmography Estimation
A two-stage framework combining flow matching and residual diffusion to refine robust rPPG waveforms.
M.S. Student · Jeonbuk National University
Juha Park
Physiological Computing & Computer Vision Researcher
Electronic & Information Engineering · advised by Prof. Sang Jun Lee
Research snapshot
How reliably can ordinary cameras recover physiological signals outside controlled settings?
I work on robust remote PPG, non-contact stress estimation, egocentric vision, and medical image analysis.
01 / RESEARCH
Research focus
My research combines computer vision, biosignal processing, and deep learning to measure physiological signals without physical contact.
Current projects cover robust remote photoplethysmography (rPPG), stress estimation, and heart-rate estimation from egocentric NIR eye video. My earlier work includes maxillary-sinus segmentation in panoramic X-rays and knowledge distillation for semantic segmentation.
Long term, I am interested in physiological sensing with egocentric and wearable cameras.
01Remote PhotoplethysmographyRecovering pulse signals from subtle facial color variation.
02Affective & Physiological ComputingInferring stress and human state from non-contact signals.
03Medical Image AnalysisDeep segmentation for dental and maxillofacial imaging.
02 / SELECTED WORK
Publications
Selected journal and conference work in physiological sensing, medical imaging, and computer vision.
Showing all 5 works.
Journal Articles
Panoramic radiographs
MedSAM
compared with
DeepLabv3+
Maxillary sinus segmentation
Deep Learning-Based Segmentation of the Maxillary Sinus on Panoramic Radiographs Using MedSAM and DeepLabv3+
A clinical comparison of MedSAM and DeepLabv3+ for automated maxillary sinus segmentation in 1,046 panoramic radiographs.
Weighted Knowledge Distillation for Semi-Supervised Segmentation of Maxillary Sinus in Panoramic X-ray Images
A semi-supervised framework that uses weighted knowledge distillation to suppress unreliable signals from teacher–student structural mismatches and SinusCycle-GAN to refine pseudo-labels.
Conference Papers
PRISSM: PRV-Guided SSM for Non-contact Stress Estimation
Pulse-rate-variability guidance meets selective state-space modeling for camera-based stress recognition.
BEAT: Bottleneck-Enhanced Adaptive Temporal Learning for Heart Rate Estimation from Egocentric NIR Eye Videos
Temporal learning for contact-free heart-rate sensing in the challenging egocentric near-infrared eye setting.
ECCV WORKSHOP03 / IN KOREAN
In Korean
Domestic journal and conference papers published or presented in Korean.
Journal Articles
Channel Attention Based Knowledge Distillation for Segmentation of Country Club Driving Environments
A lightweight student network learns channel-aware feature importance for outdoor driving-scene segmentation.
Attention-based U-Net Transformer for Segmentation of Maxillary Sinus Regions in Panoramic X-ray Images
An attention-enhanced U-Net transformer for delineating maxillary sinus regions in dental panoramic radiographs.
Conference Papers
CONFERENCE01
Channel Attention Based Knowledge Distillation for Segmentation of Country Club Driving Environments
Jeonbuk National University International Convention Center · Jun 25–27, 2025
CONFERENCE02
Maxillary Sinus Detection in Panoramic X-ray Images Using an Attention-Based U-Net Transformer
Jeju Gravel Hotel · Nov 13–16, 2024
04 / JOURNEY
Education
Building a foundation across electronics, computer vision, and physiological computing.
CERTIFICATION · 2023AI-900: Azure AI Fundamentals
Microsoft · Credential ID: GTcD-4wBm
View credential ↗MAR 2026 — PRESENT
M.S. in Electronic & Information Engineering
Jeonbuk National University · Jeonju, Republic of Korea
Advised by Prof. Sang Jun Lee. Researching robust rPPG and contact-free stress estimation.
MAR 2022 — FEB 2026
B.S. in Electronic Engineering
Jeonbuk National University · Jeonju, Republic of Korea
Developed a foundation in signal processing, embedded systems, computer vision, and deep learning.
05 / RECOGNITION
Awards & Honors
11 awards and honors in research funding, AI, engineering design, entrepreneurship, internships, and mentoring.
Representative achievements
Selected honors
Grand Prize
National Contest of Champions
Seoul National UniversityPresident’s Award
Startup Idea Challenge for Social Big Data Analysis
University of SeoulGold Prize
Creative Engineering Design Contest
Jeonbuk National UniversityShared across awards
Keywords used by 2+ awardsFive keywords are shared by two or more awards. Select one to highlight every match.
Complete record
Additional recognitions
- Master’s Student Research Encouragement GrantScience & Engineering Academic Research Support Program · Sep 2026–Aug 2027National Research Foundation of KoreaSource (opens in a new tab)
- Excellence AwardMentoring ProgramJeonbuk National UniversityCertificate (opens in a new tab)
- Outstanding Paper AwardMaxillary Sinus Detection in Panoramic X-ray ImagesIEMEK Fall ConferenceCertificate (opens in a new tab)
- Outstanding InternBig Data Innovation Convergence CollegeSeoul National UniversityCertificate (opens in a new tab)
- Merit AwardArtificial Intelligence Online CompetitionJeonbuk National UniversityCertificate (opens in a new tab)
- Merit AwardAWS DeepRacer CompetitionJeonbuk National UniversityCertificate (opens in a new tab)
- Excellence AwardSmart Device Idea ContestJeonbuk National UniversityCertificate (opens in a new tab)
- Bronze PrizeTBM Startup CampJeonbuk National UniversityCertificate (opens in a new tab)
06 / OPEN TO OPPORTUNITIES
I’m currently looking for opportunities.
I work on camera-based physiological sensing, computer vision, and medical AI. I’m exploring research and AI engineering opportunities, and I’m also happy to discuss collaborations.