arXiv subjects

Hyunwoo Kang

Publications and source records attributed to Hyunwoo Kang.

17 recordsLinked to original sources

arXiv2026

Analyzing RV32/RV64 Trade-offs for FreeRTOS Latency on 8-Stage RISC-V Soft Processors

Although many commercial RISC-V platforms provide real-time operating system support, practical examples that explain how to enable a preemptive RTOS on a custom bare-metal RISC-V soft processor remain limited, leaving the interaction between processor microarchitecture, interrupt handling, and RTOS context switching difficult to understand from simple hardware implementation examples. This paper presents the design and evaluation of FreeRTOS on custom 8-stage RV32/RV64 RISC-V soft processors with machine-mode CSRs, a CLINT timer, trap and exception control logic, and the required context-switch path. Using this platform, we compare RV32 and RV64 under the same microarchitectural organization and firmware structure using two Rhealstone-derived latency microbenchmarks, task switching and task preemption, measured with the mcycle counter. RV64 requires 36.6% and 17.7% more cycles for task switching and preemption, respectively. Instruction-level analysis attributes this overhead to the doubled RV64 trap frame, wider pointer-based kernel data structures, and 64-bit scheduler priority handling. The RTL, firmware, and benchmark code are released as open source.

cs.AR↗

arXiv2026

Ambient-robust Inverse Rendering using Active RGB-NIR Imaging

Inverse rendering aims to reconstruct geometry and reflectance of objects from images. Despite recent progress, existing methods often produces inaccurate reconstructions that are sensitive to ambient illumination conditions. Here we introduce an ambient-robust inverse rendering method enabled by active RGB-NIR imaging. Our key insight is to leverage near-infrared (NIR) flash illumination-imperceptible to human observers-to obtain stable point-light shading that is largely invariant to ambient illumination. By using multi-view RGB images illuminated by ambient light and NIR images acquired with active NIR flash illumination, we reconstruct accurate geometry and reflectance by exploiting the complementary benefits of RGB and NIR images via a three-stage inverse rendering method. To enable dense multi-view acquisition, we develop an active imaging system equipped with a RGB-NIR camera and a NIR flash mounted on a mobile base. Using this system, we collect the first multi-view RGB-NIR inverse rendering dataset captured under multiple ambient illumination conditions. Experiments demonstrate that our method outperforms prior approaches, achieving accurate geometry and reflectance estimation across multiple ambient lighting scenarios.

cs.CV↗

arXiv2026

RV-IM100: Quantifying ISA Extension, Datapath Width, and Pipeline Depth Trade-offs in RISC-V Microarchitectures

While functional RISC-V implementations are readily available in academia, controlled empirical studies that extend a single baseline architecture along multiple design axes and quantify the resulting trade-offs at each step remain scarce. This paper presents RV-IM100, a family of 10 incremental FPGA-implemented microarchitectures derived from a common 5-stage pipeline baseline, systematically varying datapath width from RV32 to RV64, instruction set from I to IM, and pipeline depth from 5 to 8 stages under controlled conditions. The I-to-IM extension produced strongly benchmark-dependent effects at the 5-stage level: CoreMark throughput more than doubled while Dhrystone throughput decreased marginally despite improved per-MHz efficiency. Within the RV32IM configuration, an iterative timing-closure methodology combined with pipeline deepening from 5 to 8 stages raised max frequency from 43 to 126MHz, increasing both Dhrystone and CoreMark throughput by 71%, while per-MHz efficiency decreased by 41%. The 6-to-7-stage transition caused throughput regression in RV64 despite higher frequency, revealing that the outcome depends on available frequency headroom. Cross-width comparison showed RV32 outperforming RV64 in absolute throughput, with per-MHz efficiency diverging by benchmark: RV64 led by 2.3% in DMIPS/MHz while RV32 led by 4.6% in CoreMark/MHz. At 8 stages, RV32 required 59% fewer LUTs, 51% fewer FFs, and 80% fewer DSPs, indicating that the resource cost of width extension substantially exceeds the modest efficiency differences. These results provide a quantitative reference for design-space exploration in RISC-V microarchitectures. All RTL sources and benchmark configurations are publicly available.

cs.AR↗

arXiv2024

Towards High-fidelity Head Blending with Chroma Keying for Industrial Applications

We introduce an industrial Head Blending pipeline for the task of seamlessly integrating an actor's head onto a target body in digital content creation. The key challenge stems from discrepancies in head shape and hair structure, which lead to unnatural boundaries and blending artifacts. Existing methods treat foreground and background as a single task, resulting in suboptimal blending quality. To address this problem, we propose CHANGER, a novel pipeline that decouples background integration from foreground blending. By utilizing chroma keying for artifact-free background generation and introducing Head shape and long Hair augmentation ($H^2$ augmentation) to simulate a wide range of head shapes and hair styles, CHANGER improves generalization on innumerable various real-world cases. Furthermore, our Foreground Predictive Attention Transformer (FPAT) module enhances foreground blending by predicting and focusing on key head and body regions. Quantitative and qualitative evaluations on benchmark datasets demonstrate that our CHANGER outperforms state-of-the-art methods, delivering high-fidelity, industrial-grade results.

cs.CV↗

arXiv2023

Design and Operation of Autonomous Wheelchair Towing Robot

In this study, a new concept of a wheelchair-towing robot for the facile electrification of manual wheelchairs is introduced. The development of this concept includes the design of towing robot hardware and an autonomous driving algorithm to ensure the safe transportation of patients to their intended destinations inside the hospital. We developed a novel docking mechanism to facilitate easy docking and separation between the towing robot and the manual wheelchair, which is connected to the front caster wheel of the manual wheelchair. The towing robot has a mecanum wheel drive, enabling the robot to move with a high degree of freedom in the standalone driving mode while adhering to kinematic constraints in the docking mode. Our novel towing robot features a camera sensor that can observe the ground ahead which allows the robot to autonomously follow color-coded wayfinding lanes installed in hospital corridors. This study introduces dedicated image processing techniques for capturing the lanes and control algorithms for effectively tracing a path to achieve autonomous path following. The autonomous towing performance of our proposed platform was validated by a real-world experiment in which a hospital environment with colored lanes was created.

cs.RO↗

arXiv2023

Renovation of Seoul Radio Astronomy Observatory and Its First Millimeter VLBI Observations

The Seoul Radio Astronomy Observatory (SRAO) operates a 6.1-meter radio telescope on the Gwanak campus of Seoul National University. We present the efforts to reform SRAO to a Very Long Baseline Interferometry (VLBI) station, motivated by recent achievements by millimeter interferometer networks such as Event Horizon Telescope, East Asia VLBI Network, and Korean VLBI Network (KVN). For this goal, we installed a receiver that had been used in the Combined Array for Research in Millimeter-wave Astronomy and a digital backend, including an H-maser clock. The existing hardware and software were also revised, which had been dedicated only to single-dish operations. After several years of preparations and test observations in 1 and 3-millimeter bands, a fringe was successfully detected toward 3C 84 in 86 GHz in June 2022 for a baseline between SRAO and KVN Ulsan station separated by 300 km. Thanks to the dual frequency operation of the receiver, the VLBI observations will soon be extended to the 1 mm band and verify the frequency phase referencing technique between 1 and 3-millimeter bands.

astro-ph.IM↗

arXiv2022

OAMixer: Object-aware Mixing Layer for Vision Transformers

Patch-based models, e.g., Vision Transformers (ViTs) and Mixers, have shown impressive results on various visual recognition tasks, alternating classic convolutional networks. While the initial patch-based models (ViTs) treated all patches equally, recent studies reveal that incorporating inductive bias like spatiality benefits the representations. However, most prior works solely focused on the location of patches, overlooking the scene structure of images. Thus, we aim to further guide the interaction of patches using the object information. Specifically, we propose OAMixer (object-aware mixing layer), which calibrates the patch mixing layers of patch-based models based on the object labels. Here, we obtain the object labels in unsupervised or weakly-supervised manners, i.e., no additional human-annotating cost is necessary. Using the object labels, OAMixer computes a reweighting mask with a learnable scale parameter that intensifies the interaction of patches containing similar objects and applies the mask to the patch mixing layers. By learning an object-centric representation, we demonstrate that OAMixer improves the classification accuracy and background robustness of various patch-based models, including ViTs, MLP-Mixers, and ConvMixers. Moreover, we show that OAMixer enhances various downstream tasks, including large-scale classification, self-supervised learning, and multi-object recognition, verifying the generic applicability of OAMixer

cs.CV↗

arXiv2021

Object-aware Contrastive Learning for Debiased Scene Representation

Contrastive self-supervised learning has shown impressive results in learning visual representations from unlabeled images by enforcing invariance against different data augmentations. However, the learned representations are often contextually biased to the spurious scene correlations of different objects or object and background, which may harm their generalization on the downstream tasks. To tackle the issue, we develop a novel object-aware contrastive learning framework that first (a) localizes objects in a self-supervised manner and then (b) debias scene correlations via appropriate data augmentations considering the inferred object locations. For (a), we propose the contrastive class activation map (ContraCAM), which finds the most discriminative regions (e.g., objects) in the image compared to the other images using the contrastively trained models. We further improve the ContraCAM to detect multiple objects and entire shapes via an iterative refinement procedure. For (b), we introduce two data augmentations based on ContraCAM, object-aware random crop and background mixup, which reduce contextual and background biases during contrastive self-supervised learning, respectively. Our experiments demonstrate the effectiveness of our representation learning framework, particularly when trained under multi-object images or evaluated under the background (and distribution) shifted images.

cs.CV↗

arXiv2021

TRAO Survey of the nearby filamentary molecular clouds, the universal nursery of stars (TRAO FUNS). II. Filaments and Dense cores in IC 5146

We present the results on the physical properties of filaments and dense cores in IC 5146, as a part of the TRAO FUNS project. We carried out On-The-Fly mapping observations using the Taeduk Radio Astronomy Observatory (TRAO) 14m telescope covering about 1 square degree of the area of IC 5146 using various molecular lines. We identified 14 filaments (24 in total, including sub-filaments) from the C$^{18}$O (1-0) data cube and 22 dense cores from the $\rm N_{2}H^{+}$ (1-0) data. We examined the filaments' gravitational criticality, turbulence properties, accretion rate from filaments to dense cores, and relative evolutionary stages of cores. Most filaments in IC 5146 are gravitationally supercritical within the uncertainty, and most dense cores are formed in them. We found that dense cores in the hubs show a systemic velocity shift of ~0.3 km/s between the $\rm N_{2}H^{+}$ and C$^{18}$O gas. Besides, these cores are subsonic or transonic, while the surrounding filament gas is transonic or supersonic, indicating that the cores in the hubs are likely formed by the turbulence dissipation in the colliding turbulent filaments and the merging is still ongoing. We estimated the mass accretion rate of $15 - 35~M_{\odot}~\rm Myr^{-1}$ from the filaments to the dense cores, and the required time scales to collect the current core mass are consistent with the lifetime of the dense cores. The structures of filaments and dense cores in the hub can form by a collision of turbulent converging flows, and mass flow along the filaments to the dense cores may play an important role in forming dense cores.

astro-ph.GA↗

arXiv2021

TIMES II: Investigating the Relation Between Turbulence and Star-forming Environments in Molecular Clouds

We investigate the effect of star formation on turbulence in the Orion A and Ophiuchus clouds using principal component analysis (PCA). We measure the properties of turbulence by applying PCA on the spectral maps in $^{13}$CO, C$^{18}$O, HCO$^+$ $J=$1$-$0, and CS $J=$2$-$1. First, the scaling relations derived from PCA of the $^{13}$CO maps show that the velocity difference ($δv$) for a given spatial scale ($L$) is the highest in the integral shaped filament (ISF) and L1688, where the most active star formation occurs in the two clouds. The $δv$ increases with the number density and total bolometric luminosity of the protostars in the sub-regions. Second, in the ISF and L1688 regions, the $δv$ of C$^{18}$O, HCO$^+$, and CS are generally higher than that of $^{13}$CO, which implies that the dense gas is more turbulent than the diffuse gas in the star-forming regions; stars form in dense gas, and dynamical activities associated with star formation, such as jets and outflows, can provide energy into the surrounding gas to enhance turbulent motions.

astro-ph.GA↗

arXiv2021

TIMES I: a Systematic Observation in Multiple Molecular Lines Toward the Orion A and Ophiuchus Clouds

We have used the Taeduk Radio Astronomy Observatory to observe the Orion A and Ophiuchus clouds in the $J=$1$-$0 lines of $^{13}$CO, C$^{18}$O, HCN, HCO$^+$, and N$_2$H$^+$ and the $J=$2$-$1 line of CS. The fully sampled maps with uniform noise levels are used to create moment maps. The variations of the line intensity and velocity dispersion with total column density, derived from dust emission maps, are presented and compared to previous work. The CS line traces dust column density over more than one order of magnitude, and the N$_2$H$^+$ line best traces the highest column density regime ($\log(N_\mathrm{H_2}$) $>$ 22.8). Line luminosities, integrated over the cloud, are compared to those seen in other galaxies. The HCO$^+$-to-HCN luminosity ratio in the Orion A cloud is similar to that of starburst galaxies, while that in the Ophiuchus cloud is in between those of active galactic nuclei and starburst galaxies.

astro-ph.GA↗

arXiv2020

Cloud structures in M17 SWex : Possible cloud-cloud collision

Using wide-field $^{13}$CO ($J=1-0$) data taken with the Nobeyama 45-m telescope, we investigate cloud structures of the infrared dark cloud complex in M17 with SCIMES. In total, we identified 118 clouds that contain 11 large clouds with radii larger than 1 pc. The clouds are mainly distributed in the two representative velocity ranges of 10 $-$ 20 km s$^{-1}$ and 30 $-$ 40 km s$^{-1}$. By comparing with the ATLASGAL catalog, we found that the majority of the $^{13}$CO clouds with 10 $-$ 20 km s$^{-1}$ and 30 $-$ 40 km s$^{-1}$ are likely located at distances of 2 kpc (Sagittarius arm) and 3 kpc (Scutum arm), respectively. Analyzing the spatial configuration of the identified clouds and their velocity structures, we attempt to reveal the origin of the cloud structure in this region. Here we discuss three possibilities: (1) overlapping with different velocities, (2) cloud oscillation, and (3) cloud-cloud collision. From the position-velocity diagrams, we found spatially-extended faint emission between $\sim$ 20 km s$^{-1}$ and $\sim$ 35 km s$^{-1}$, which is mainly distributed in the spatially-overlapped areas of the clouds. We also found that in some areas where clouds with different velocities overlapped, the magnetic field orientation changes abruptly. The distribution of the diffuse emission in the position-position-velocity space and the bending magnetic fields appear to favor the cloud-cloud collision scenario compared to other scenarios. In the cloud-cloud collision scenario, we propose that two $\sim$35 km s$^{-1}$ foreground clouds are colliding with clouds at $\sim$20 km s$^{-1}$ with a relative velocity of 15 km s$^{-1}$. These clouds may be substructures of two larger clouds having velocities of $\sim$ 35 km s$^{-1}$ ($\gtrsim 10^3 $ M$_{\odot}$) and $\sim$ 20 km s$^{-1}$ ($\gtrsim 10^4 $ M$_{\odot}$), respectively.

astro-ph.GA↗

arXiv2019

TRAO Survey of Nearby Filamentary Molecular clouds, the Universal Nursery of Stars (TRAO FUNS) I. Dynamics and Chemistry of L1478 in the California Molecular Cloud

"TRAO FUNS" is a project to survey Gould Belt's clouds in molecular lines. This paper presents its first results on the central region of the California molecular cloud, L1478. We performed On-The-Fly mapping observations using the Taedeok Radio Astronomy Observatory (TRAO) 14m single dish telescope equipped with a 16 multi-beam array covering $\sim$1.0 square degree area of this region using C$^{18}$O (1-0) mainly tracing low density cloud and about 460 square arcminute area using N$_{2}$H$^{+}$ (1-0) mainly tracing dense cores. CS (2-1) and SO $(3_{2}-2_{1})$ were also used simultaneously to map $\sim$440 square arcminute area of this region. We identified 10 filaments by applying the dendrogram technique to the C$^{18}$O data-cube and 8 dense N$_{2}$H$^{+}$ cores by using {\sc FellWalker}. Basic physical properties of filaments such as mass, length, width, velocity field, and velocity dispersion are derived. It is found that L1478 consists of several filaments with slightly different velocities. Especially the filaments which are supercritical are found to contain dense cores detected in N$_{2}$H$^{+}$. Comparison of non-thermal velocity dispersions derived from C$^{18}$O and N$_{2}$H$^{+}$ for the filaments and dense cores indicates that some of dense cores share similar kinematics with those of the surrounding filaments while several dense cores have different kinematics with those of their filaments. This suggests that the formation mechanism of dense cores and filaments can be different in individual filaments depending on their morphologies and environments.

astro-ph.GA↗

arXiv2018

The dense galactic environments of the Milky Way

Star formation takes place in the dense gas phase, and therefore a simple dense gas and star formation rate relation has been proposed. With the advent of multi-beam receivers, new observations show that the deviation from linear relations is possible. In addition, different dense gas tracers might also change significantly the measurement of dense gas mass and subsequently the relation between star formation rate and dense gas mass. We report the preliminary results the DEnse GAs in MAssive star-forming regions in the Milky Way (DEGAMA) survey that observed the dense gas toward a suit of well-characterized massive star forming regions in the Milky Way. Using the resulting maps of HCO$^{+}$ 1--0, HCN 1--0, CS 2--1, we discuss the current understanding of the dense gas phase where star formation takes place.

astro-ph.GA↗

arXiv2015

Simultaneous observation of water and class I methanol masers toward class II methanol maser sources

We present a simultaneous single-dish survey of 22 GHz water maser and 44 GHz and 95 GHz class I methanol masers toward 77 6.7 GHz class II methanol maser sources, which were selected from the Arecibo methanol maser Galactic plane survey (AMGPS) catalog.Water maser emission is detected in 39 (51%) sources, of which 15 are new detections. Methanol maser emission at 44 GHz and 95 GHz is found in 25 (32%) and 19 (25%) sources, of which 21 and 13 sources are newly detected, respectively. We find 4 high-velocity (> 30 km/s) water maser sources, including 3 dominant blue- or redshifted outflows.The 95 GHz masers always appear with the 44 GHz maser emission. They are strongly correlated with 44 GHz masers in velocity, flux density, and luminosity, while they are not correlated with either water or 6.7 GHz class II methanol masers. The average peak flux density ratio of 95 GHz to 44 GHz masers is close to unity, which is two times higher than previous estimates. The flux densities of class I methanol masers are more closely correlated with the associated BGPS core mass than those of water or class II methanol masers. Using the large velocity gradient (LVG) model and assuming unsaturated class I methanol maser emission, we derive the fractional abundance of methanol to be in a range of 4.2*10^-8 to 2.3*10^-6, with a median value of 3.3\pm2.7*10^-7.

astro-ph.SR↗

arXiv2012

SRAO CO Observation of 11 Supernova Remnants in l = 70 to 190 deg

We present the results of 12CO J = 1-0 line observations of eleven Galactic supernova remnants (SNRs) obtained using the Seoul Radio Astronomy Observatory (SRAO) 6-m radio telescope. The observation was made as a part of the SRAO CO survey of SNRs between l = 70 and 190 deg, which is intended to identify SNRs interacting with molecular clouds. The mapping areas for the individual SNRs are determined to cover their full extent in the radio continuum. We used halfbeam grid spacing (60") for 9 SNRs and full-beam grid spacing (120") for the rest. We detected CO emission towards most of the remnants. In six SNRs, molecular clouds showed a good spatial relation with their radio morphology, although no direct evidence for the interaction was detected. Two SNRs are particularly interesting: G85.4+0.7, where there is a filamentary molecular cloud along the radio shell, and 3C434.1, where a large molecular cloud appears to block the western half of the remnant. We briefly summarize the results obtained for individual SNRs.

astro-ph.GA↗

arXiv2011

A Multi-Epoch, Simultaneous Water and Methanol Maser Survey toward Intermediate-Mass Young Stellar Objects

We report a multi-epoch, simultaneous 22 GHz H2O and 44 GHz class I CH3OH maser line survey towards 180 intermediate-mass young stellar objects, including 14 Class 0, 19 Class I objects, and 147 Herbig Ae/Be stars. We detected H2O and CH3OH maser emission towards 16 (9%) and 10 (6%) sources with one new H2O and six new CH3OH maser sources. The detection rates of both masers rapidly decrease as the central (proto)stars evolve, which is contrary to the trends in high-mass star-forming regions. This suggests that the excitations of the two masers are closely related to the evolutionary stage of the central (proto)stars and the circumstellar environments. H2O maser velocities deviate on average 9 km s^-1 from the ambient gas velocities whereas CH3OH maser velocities match quite well with the ambient gas velocities. For both maser emissions, large velocity differences (|v_{H2O} - v_{sys} | > 10 km s^-1 and |v_{CH3OH} - v_{sys}| > 1 km s^-1) are mostly confined to Class 0 objects. The formation and disappearance of H2O masers is frequent and their integrated intensities change by up to two orders of magnitude. In contrast, CH3OH maser lines usually show no significant change in intensity, shape, or velocity. This is consistent with the previous suggestion that H2O maser emission originates from the base of an outflow while 44 GHz class I CH3OH maser emission arises from the interaction region of the outflow with the ambient gas. The isotropic maser luminosities are well correlated with the bolometric luminosities of the central objects. The fitted relations are L_{H2O} = 1.71 * 10^{-9} (L_{bol})^{0.97} and L_{CH3OH} = 1.71 * 10^{-10} (L_{bol})^{1.22}.

astro-ph.GA↗