Showing 1–9 of 9 results for author: Ki, S

CLIP-Guided Attribute Aware Pretraining for Generalizable Image Quality Assessment

Authors: Daekyu Kwon, Dongyoung Kim, Sehwan Ki, Younghyun Jo, Hyong-Euk Lee, Seon Joo Kim

Abstract: In no-reference image quality assessment (NR-IQA), the challenge of limited dataset sizes hampers the development of robust and generalizable models. Conventional methods address this issue by utilizing large datasets to extract rich representations for IQA. Also, some approaches propose vision language models (VLM) based IQA, but the domain gap between generic VLM and IQA constrains their scalabi… ▽ More In no-reference image quality assessment (NR-IQA), the challenge of limited dataset sizes hampers the development of robust and generalizable models. Conventional methods address this issue by utilizing large datasets to extract rich representations for IQA. Also, some approaches propose vision language models (VLM) based IQA, but the domain gap between generic VLM and IQA constrains their scalability. In this work, we propose a novel pretraining framework that constructs a generalizable representation for IQA by selectively extracting quality-related knowledge from VLM and leveraging the scalability of large datasets. Specifically, we carefully select optimal text prompts for five representative image quality attributes and use VLM to generate pseudo-labels. Numerous attribute-aware pseudo-labels can be generated with large image datasets, allowing our IQA model to learn rich representations about image quality. Our approach achieves state-of-the-art performance on multiple IQA datasets and exhibits remarkable generalization capabilities. Leveraging these strengths, we propose several applications, such as evaluating image generation models and training image enhancement models, demonstrating our model's real-world applicability. We will make the code available for access. △ Less

Submitted 3 June, 2024; originally announced June 2024.

Information-Theoretic GAN Compression with Variational Energy-based Model

Authors: Minsoo Kang, Hyewon Yoo, Eunhee Kang, Sehwan Ki, Hyong-Euk Lee, Bohyung Han

Abstract: We propose an information-theoretic knowledge distillation approach for the compression of generative adversarial networks, which aims to maximize the mutual information between teacher and student networks via a variational optimization based on an energy-based model. Because the direct computation of the mutual information in continuous domains is intractable, our approach alternatively optimize… ▽ More We propose an information-theoretic knowledge distillation approach for the compression of generative adversarial networks, which aims to maximize the mutual information between teacher and student networks via a variational optimization based on an energy-based model. Because the direct computation of the mutual information in continuous domains is intractable, our approach alternatively optimizes the student network by maximizing the variational lower bound of the mutual information. To achieve a tight lower bound, we introduce an energy-based model relying on a deep neural network to represent a flexible variational distribution that deals with high-dimensional images and consider spatial dependencies between pixels, effectively. Since the proposed method is a generic optimization algorithm, it can be conveniently incorporated into arbitrary generative adversarial networks and even dense prediction networks, e.g., image enhancement models. We demonstrate that the proposed algorithm achieves outstanding performance in model compression of generative adversarial networks consistently when combined with several existing models. △ Less

Submitted 28 March, 2023; originally announced March 2023.

Comments: Accepted at Neurips2022

Transverse Magnetic Mode Laser in Photonic Crystal Nanobeam Cavity

Authors: Taesu Ryu, Hwi-Min Kim, Sang-Woo Ki, Yong-Hee Lee, **-Kyu Yang

Abstract: We first experimentally demonstrated a transverse magnetic (TM) mode laser in a Photonic crystal (PhC) slab structure at room temperature. This study proposes a PhC nanobeam (NB) cavity to support a high-quality-factor (Q-factor) TM mode. For a large and complete photonic bandgap, the PhC NB structures consist of large air holes in a thick dielectric slab. The PhC NB cavity was optimized numerical… ▽ More We first experimentally demonstrated a transverse magnetic (TM) mode laser in a Photonic crystal (PhC) slab structure at room temperature. This study proposes a PhC nanobeam (NB) cavity to support a high-quality-factor (Q-factor) TM mode. For a large and complete photonic bandgap, the PhC NB structures consist of large air holes in a thick dielectric slab. The PhC NB cavity was optimized numerically for a high-Q-factor TM mode of over 1,000,000 by reducing the radii of the air holes quadratically from the center to the edge of the PhC NB. A single TM mode lasing action was observed in an InGaAsP quantum-well (QW)-embedded optimized PhC NB cavity structure at room temperature via optical pulse pum**, where the QW layer was lightly etched. We believe that the TM mode lasers in PhC NB cavities with a lightly etched QW can be good candidates for a surface plasmon excitation source or a highly sensitive optical sensor. △ Less

Submitted 21 December, 2022; originally announced December 2022.

Automated coordination corrected enthalpies with AFLOW-CCE

Authors: R. Friedrich, M. Esters, C. Oses, S. Ki, M. J. Brenner, D. Hicks, M. J. Mehl, C. Toher, S. Curtarolo

Abstract: The computational design of materials with ionic bonds poses a critical challenge to thermodynamic modeling since density functional theory yields inaccurate predictions of their formation enthalpies. Progress requires leveraging physically insightful correction methods. The recently introduced coordination corrected enthalpies (CCE) method delivers accurate formation enthalpies with mean absolute… ▽ More The computational design of materials with ionic bonds poses a critical challenge to thermodynamic modeling since density functional theory yields inaccurate predictions of their formation enthalpies. Progress requires leveraging physically insightful correction methods. The recently introduced coordination corrected enthalpies (CCE) method delivers accurate formation enthalpies with mean absolute errors close to room temperature thermal energy, i.e., 25meV/atom. The CCE scheme, depending on the number of cation-anion bonds and oxidation state of the cation, requires an automated analysis of the system to determine and apply the correction. Here, we present AFLOW-CCE -- our implementation of CCE into the AFLOW framework for computational materials design. It features a command line tool, a web interface and a Python environment. The workflow includes a structural analysis, automatically determines oxidation numbers, and accounts for temperature effects by parametrizing vibrational contributions to the formation enthalpy per bond. △ Less

Submitted 12 March, 2021; v1 submitted 7 January, 2021; originally announced January 2021.

Comments: 11 pages, 4 figures, 2 tables

Journal ref: Phys. Rev. Materials 5, 043803 (2021)

A Novel Just-Noticeable-Difference-based Saliency-Channel Attention Residual Network for Full-Reference Image Quality Predictions

Authors: Soomin Seo, Sehwan Ki, Munchurl Kim

Abstract: Recently, due to the strength of deep convolutional neural networks (CNN), many CNN-based image quality assessment (IQA) models have been studied. However, previous CNN-based IQA models likely have yet to utilize the characteristics of the human visual system (HVS) fully for IQA problems when they simply entrust everything to the CNN, expecting it to learn from a training dataset. However, in this… ▽ More Recently, due to the strength of deep convolutional neural networks (CNN), many CNN-based image quality assessment (IQA) models have been studied. However, previous CNN-based IQA models likely have yet to utilize the characteristics of the human visual system (HVS) fully for IQA problems when they simply entrust everything to the CNN, expecting it to learn from a training dataset. However, in this paper, we propose a novel saliency-channel attention residual network based on the just-noticeable-difference (JND) concept for full-reference image quality assessments (FR-IQA). It is referred to as JND-SalCAR and shows significant improvements in large IQA datasets with various types of distortion. The proposed JND-SalCAR effectively learns how to incorporate human psychophysical characteristics, such as visual saliency and JND, into image quality predictions. In the proposed network, a SalCAR block is devised so that perceptually important features can be extracted with the help of saliency-based spatial attention and channel attention schemes. In addition, a saliency map serves as a guideline for predicting a patch weight map in order to afford stable training of end-to-end optimization for the JND-SalCAR. To the best of our knowledge, our work presents the first HVS-inspired trainable FR-IQA network that considers both visual saliency and the JND characteristics of the HVS. When the visual saliency map and the JND probability map are explicitly given as priors, they can be usefully combined to predict IQA scores rated by humans more precisely, eventually leading to performance improvements and faster convergence. The experimental results show that the proposed JND-SalCAR significantly outperforms all recent state-of-the-art FR-IQA methods on large IQA datasets in terms of the Spearman rank order coefficient (SRCC) and the Pearson linear correlation coefficient (PLCC). △ Less

Submitted 16 October, 2020; v1 submitted 14 February, 2019; originally announced February 2019.

Comments: Accepted for publication in IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)

Measurement of the average shape of longitudinal profiles of cosmic-ray air showers at the Pierre Auger Observatory

Authors: The Pierre Auger Collaboration, A. Aab, P. Abreu, M. Aglietta, I. F. M. Albuquerque, J. M. Albury, I. Allekotte, A. Almela, J. Alvarez Castillo, J. Alvar ez-Muñiz, G. A. Anastasi, L. Anchordoqui, B. Andrada, S. Andringa, C. Aramo, H. Asorey, P. Assis, G. Avila, A. M. Badescu, A. Bakalova, A. Balaceanu, F. Barbato, R. J. Barreira Luz, S. Baur, K. H. Becker , et al. (363 additional authors not shown)

Abstract: The profile of the longitudinal development of showers produced by ultra-high energy cosmic rays carries information related to the interaction properties of the primary particles with atmospheric nuclei. In this work, we present the first measurement of the average shower profile in traversed atmospheric depth at the Pierre Auger Observatory. The shapes of profiles are well reproduced by the Gais… ▽ More The profile of the longitudinal development of showers produced by ultra-high energy cosmic rays carries information related to the interaction properties of the primary particles with atmospheric nuclei. In this work, we present the first measurement of the average shower profile in traversed atmospheric depth at the Pierre Auger Observatory. The shapes of profiles are well reproduced by the Gaisser-Hillas parametrization within the range studied, for E > 10^{17.8} eV. A detailed analysis of the systematic uncertainties is performed using 10 years of data and a full detector simulation. The average shape is quantified using two variables related to the width and asymmetry of the profile, and the results are compared with predictions of hadronic interaction models for different primary particles. △ Less

Submitted 16 May, 2019; v1 submitted 12 November, 2018; originally announced November 2018.

Comments: Published version, 20 pages, 6 figures, 2 tables

Report number: FERMILAB-PUB-18-633

Journal ref: JCAP03(2019)018

COHERENT Collaboration data release from the first observation of coherent elastic neutrino-nucleus scattering

Authors: COHERENT Collaboration, D. Akimov, J. B. Albert, P. An, C. Awe, P. S. Barbeau, B. Becker, V. Belov, M. A. Blackston, A. Bolozdynya, A. Brown, A. Burenkov, B. Cabrera-Palmer, M. Cervantes, J. I. Collar, R. J. Cooper, R. L. Cooper, C. Cuesta, J. Daughhetee, D. J. Dean, M. del Valle Coello, J. Detwiler, M. D'Onofrio, A. Eberhardt, Y. Efremenko , et al. (69 additional authors not shown)

Abstract: This release includes data and information necessary to perform independent analyses of the COHERENT result presented in Akimov et al., arXiv:1708.01294 [nucl-ex]. Data is shared in a binned, text-based format, including both "signal" and "background" regions, so that counts and associated uncertainties can be quantitatively calculated for the purpose of separate analyses. This document describes… ▽ More This release includes data and information necessary to perform independent analyses of the COHERENT result presented in Akimov et al., arXiv:1708.01294 [nucl-ex]. Data is shared in a binned, text-based format, including both "signal" and "background" regions, so that counts and associated uncertainties can be quantitatively calculated for the purpose of separate analyses. This document describes the included information and its format, offering some guidance on use of the data. Accompanying code examples show basic interaction with the data using Python. △ Less

Submitted 25 April, 2018; originally announced April 2018.

COHERENT 2018 at the Spallation Neutron Source

Authors: D. Akimov, J. B. Albert, P. An, C. Awe, P. S. Barbeau, B. Becker, V. Belov, M. A. Blackston, A. Bolozdynya, A. Brown, A. Burenkov, B. Cabrera-Palmer, M. Cervantes, J. I. Collar, R. J. Cooper, R. L. Cooper, J. Daughhetee, D. J. Dean, M. del Valle Coello, J. A. Detwiler, M. D'Onofrio, Y. Efremenko, S. R. Elliott, E. Erkela, A. Etenko , et al. (54 additional authors not shown)

Abstract: The primary goal of the COHERENT collaboration is to measure and study coherent elastic neutrino-nucleus scattering (CEvNS) using the high-power, few-tens-of-MeV, pulsed source of neutrinos provided by the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). The COHERENT collaboration reported the first detection of CEvNS [Akimov:2017ade] using a CsI[Na] detector. At present th… ▽ More The primary goal of the COHERENT collaboration is to measure and study coherent elastic neutrino-nucleus scattering (CEvNS) using the high-power, few-tens-of-MeV, pulsed source of neutrinos provided by the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). The COHERENT collaboration reported the first detection of CEvNS [Akimov:2017ade] using a CsI[Na] detector. At present the collaboration is deploying four detector technologies: a CsI[Na] scintillating crystal, p-type point-contact germanium detectors, single-phase liquid argon, and NaI[Tl] crystals. All detectors are located in the neutron-quiet basement of the SNS target building at distances 20-30 m from the SNS neutrino source. The simultaneous measurement in all four COHERENT detector subsystems will test the $N^2$ dependence of the cross section and search for new physics. In addition, COHERENT is measuring neutrino-induced neutrons from charged- and neutral-current neutrino interactions on nuclei in shielding materials, which represent a non-negligible background for CEvNS as well as being of intrinsic interest. The Collaboration is planning as well to look for charged-current interactions of relevance to supernova and weak-interaction physics. This document describes concisely the COHERENT physics motivations, sensitivity, and next plans for measurements at the SNS to be accomplished on a few-year timescale. △ Less

Submitted 2 April, 2018; v1 submitted 24 March, 2018; originally announced March 2018.

Comments: 22 pages, 14 figures

Observation of Coherent Elastic Neutrino-Nucleus Scattering

Authors: D. Akimov, J. B. Albert, P. An, C. Awe, P. S. Barbeau, B. Becker, V. Belov, A. Brown, A. Bolozdynya, B. Cabrera-Palmer, M. Cervantes, J. I. Collar, R. J. Cooper, R. L. Cooper, C. Cuesta, D. J. Dean, J. A. Detwiler, A. Eberhardt, Y. Efremenko, S. R. Elliott, E. M. Erkela, L. Fabris, M. Febbraro, N. E. Fields, W. Fox , et al. (56 additional authors not shown)

Abstract: The coherent elastic scattering of neutrinos off nuclei has eluded detection for four decades, even though its predicted cross-section is the largest by far of all low-energy neutrino couplings. This mode of interaction provides new opportunities to study neutrino properties, and leads to a miniaturization of detector size, with potential technological applications. We observe this process at a 6.… ▽ More The coherent elastic scattering of neutrinos off nuclei has eluded detection for four decades, even though its predicted cross-section is the largest by far of all low-energy neutrino couplings. This mode of interaction provides new opportunities to study neutrino properties, and leads to a miniaturization of detector size, with potential technological applications. We observe this process at a 6.7-sigma confidence level, using a low-background, 14.6-kg CsI[Na] scintillator exposed to the neutrino emissions from the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory. Characteristic signatures in energy and time, predicted by the Standard Model for this process, are observed in high signal-to-background conditions. Improved constraints on non-standard neutrino interactions with quarks are derived from this initial dataset. △ Less

Submitted 3 August, 2017; originally announced August 2017.