Author name cluster

Jun-Hee Kim

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

1 author row

AAAI Conference 2025 Conference Paper

ProPose: Probabilistic 3D Human Pose Estimation with Instance-Level Distribution and Normalizing Flow

Jumin Han
Jun-Hee Kim
Seong-Whan Lee

3D Human Pose Estimation (HPE) is a one-to-many problem by nature, making it challenging to estimate an accurate 3D pose from a single 2D pose. Some prior works have attempted to tackle this problem by using a conditional generative network. They generate 3D poses from a given 2D pose with noises from a standard Gaussian distribution, while the depth distribution is dependent on each posture and more complex than the standard Gaussian distribution. This may lead to inaccurate distribution learning. In this paper, we propose a probabilistic framework called ProPose to address this issue. ProPose employs Pose Instance-Level Gaussian Distribution (PILGD) derived from 3D pose-based self-representation learning to obtain reliable distribution which is able to address pose-dependent depth distribution. To access this PILGD, we utilize normalizing flow, which learns a mapping function between the PILGD and a 2D Pose-Adaptive Gaussian Distribution (PAGD). This converts the problem of directly estimating 3D poses from 2D poses to a mapping problem between PILGD and PAGD using a normalizing flow. Extensive experiments show the advantages of utilizing the PILGD and PAGD. ProPose achieves comparable performances to previous state-of-the-art probabilistic methods in a multi-hypothesis setting. Notably, ProPose in a single-hypothesis setting demonstrates comparable generalization ability to existing state-of-the-art deterministic methods.

PDF Details DOI

YNIMG Journal 2025 Journal Article

Simultaneous measurement of cerebral blood flow and cerebrospinal fluid flow using pseudo-continuous arterial spin labeling

Jae-Geun Im
Jun-Hee Kim
Sung-Hong Park

In the brain clearance system, the movement of cerebrospinal fluid (CSF) plays a key role in processing waste products. Previous studies have shown that CSF flow interacts significantly with cerebral blood flow (CBF) during brain waste clearance, but there are no simultaneous measurements and comparisons of these two metrics in humans. This study introduces a novel method for simultaneously measuring CSF pulsatile movement and CBF using pseudo-continuous arterial spin labeling (pCASL) MRI. We conducted a comparative analysis of the correlation between CBF and CSF pulsatile movement in human subjects during breath-holding and motor task conditions. Our findings demonstrate the effectiveness of our proposed technique in measuring CSF pulsatile movement, as validated by comparing results with phase-contrast MRI at corresponding locations. Importantly, we observed a robust positive correlation between CBF and CSF pulsation concurrently measured through pCASL during breath-holding. Furthermore, through inter-subject comparisons of regional CBF and CSF pulsation, we established that higher blood perfusion in putamen, caudate, and pallidum regions, which are included in basal ganglia structure, corresponds to greater CSF pulsatile movement. Our motor tasks significantly increased CBF in the motor cortex, and CSF pulsation measured in the dorsal part around cisterna magna showed a decreasing tendency in the motor condition compared to the resting state, aligning with the Monroe-Kelly doctrine. Accordingly, these results demonstrate the feasibility of simultaneous measurement of CBF and CSF pulsation using the proposed pCASL technique in humans, which warrants further investigation.

Details DOI

NeurIPS Conference 2024 Conference Paper

Toward Approaches to Scalability in 3D Human Pose Estimation

Jun-Hee Kim
Seong-Whan Lee

In the field of 3D Human Pose Estimation (HPE), scalability and generalization across diverse real-world scenarios remain significant challenges. This paper addresses two key bottlenecks to scalability: limited data diversity caused by 'popularity bias' and increased 'one-to-many' depth ambiguity arising from greater pose diversity. We introduce the Biomechanical Pose Generator (BPG), which leverages biomechanical principles, specifically the normal range of motion, to autonomously generate a wide array of plausible 3D poses without relying on a source dataset, thus overcoming the restrictions of popularity bias. To address depth ambiguity, we propose the Binary Depth Coordinates (BDC), which simplifies depth estimation into a binary classification of joint positions (front or back). This method decomposes a 3D pose into three core elements—2D pose, bone length, and binary depth decision—substantially reducing depth ambiguity and enhancing model robustness and accuracy, particularly in complex poses. Our results demonstrate that these approaches increase the diversity and volume of pose data while consistently achieving performance gains, even amid the complexities introduced by increased pose diversity.

PDF Details DOI

YNIMG Journal 2022 Journal Article

Measurement of CSF pulsation from EPI-based human fMRI

Jun-Hee Kim
Jae-Geun Im
Sung-Hong Park

It is recently discovered that the glymphatic system and meningeal lymphatic system are the primary routes for the clearance of brain waste products. The CSF flow is part of these systems, facilitating the clearance procedure. Nonetheless, the relationship between CSF flow and brain functional activity has been underexplored. To investigate CSF dynamics and functional brain activity simultaneously, recent studies have proposed a CSF inflow index measured on edge slices (CSFedge) of echo-planar imaging (EPI) based functional magnetic resonance imaging (fMRI), however, it lacks the quantitative aspect of the CSF pulsation. We proposed a new method for quantifying CSF pulsation (CSFpulse) based on an interslice CSF pulsation model in the 4th ventricle of EPI-based fMRI. The proposed CSFpulse successfully detected the higher CSF flow during the resting state than the typical task states (visual and motor) ( p <. 05 ), which is consistent with previous studies based on phase contrast (PC) MRI and CSF volume MRI, while it was not detected in CSFedge based indices or baseline CSF signals in various regions of interest (ROIs). Moreover, CSFpulse demonstrated dynamic functional changes in CSF pulsation: it decreased during the activation-on blocks while it increased during the activation-off blocks. CSFpulse significantly correlated with stroke volume measured using PC MRI, a standard method for CSF pulsation quantification, under the same functional state, while CSFedge based indices or CSF ROIs showed no correlation with the PC MRI stroke volume. Lastly, the correlation of CSFpulse with global BOLD was weaker than that of CSFedge, suggesting that CSFpulse may reflect distinct CSF physiological information that is less affected by global BOLD effects. Based on these results, the proposed CSFpulse provides CSF pulsatility information more accurately in a quantitative manner than CSFedge based indices from the recent CSF studies or the conventional ROI-based analysis. In addition to the high correlation with PC MRI, CSFpulse is much faster than PC MRI and provides information of functional brain activations simultaneously, advantageous over PC MRI or CSF volume MRI. Accordingly, the suggested CSFpulse can be used for investigating intra-subject functional changes in BOLD and CSF pulsation simultaneously and inter-subject CSF pulsation variations based on conventional EPI-based fMRI, which warrants further investigation.

Details DOI