Search

Open-vocabulary object detection (OVD), detecting specific classes of objects using only their linguistic descriptions (e.g., class names) without any image samples, has garnered significant attention. However, in real-world applications, the target class concepts is often hard to describe in text and the only way to specify target objects is to provide their image examples, yet it is often challenging to obtain a good number of samples. Thus, there is a high demand from practitioners for few-shot object detection (FSOD). A natural question arises: Can the benefits of OVD extend to FSOD for object classes that are difficult to describe in text? Compared to traditional methods that learn only predefined classes (referred to in this paper as closed-set object detection, COD), can the extra cost of OVD be justified? To answer these questions, we propose a method to quantify the ``text-describability'' of object detection datasets using the zero-shot image classification accuracy with CLIP. This allows us to categorize various OD datasets with different text-describability and emprically evaluate the FSOD performance of OVD and COD methods within each category. Our findings reveal that: i) there is little difference between OVD and COD for object classes with low text-describability under equal conditions in OD pretraining; and ii) although OVD can learn from more diverse data than OD-specific data, thereby increasing the volume of training data, it can be counterproductive for classes with low-text-describability. These findings provide practitioners with valuable guidance amidst the recent advancements of OVD methods., Comment: 20 pages, 3 figures

A search for proton decay into $e^+/\mu^+$ and a $\eta$ meson has been performed using data from a 0.373 Mton$\cdot$year exposure (6050.3 live days) of Super-Kamiokande. Compared to previous searches this work introduces an improved model of the intranuclear $\eta$ interaction cross section, resulting in a factor of two reduction in uncertainties from this source and $\sim$10\% increase in signal efficiency. No significant data excess was found above the expected number of atmospheric neutrino background events resulting in no indication of proton decay into either mode. Lower limits on the proton partial lifetime of $1.4\times\mathrm{10^{34}~years}$ for $p\rightarrow e^+\eta$ and $7.3\times\mathrm{10^{33}~years}$ for $p\rightarrow \mu^+\eta$ at the 90$\%$ C.L. were set. These limits are around 1.5 times longer than our previous study and are the most stringent to date.

Diffusion models have demonstrated impressive image generation capabilities. Personalized approaches, such as textual inversion and Dreambooth, enhance model individualization using specific images. These methods enable generating images of specific objects based on diverse textual contexts. Our proposed approach aims to retain the model's original knowledge during new information integration, resulting in superior outcomes while necessitating less training time compared to Dreambooth and textual inversion.

Using SU(3) lattice QCD, we study for the first time the effective potential of the Polyakov loop $\langle P \rangle$ at finite temperature, i.e., the thermodynamic potential, in the field-theoretical way. In the framework of the reweighting method in lattice QCD, we express the effective potential $V_{\rm eff}(\langle P \rangle)$ using the expectation value without a source term. In particular, we consider the most difficult and interesting case of vacuum coexistence at the critical temperature $T_c$. We adopt SU(3) quenched lattice QCD on $48^3 \times 6$ at $\beta$= 5.89379, which corresponds exactly to the critical temperature $T_c$ of the deconfinement phase transition, and use 200,000 Monte Carlo configurations. After categorizing the gauge configurations into one $Z_3$-symmetric and three $Z_3$-broken vacua each, we perform a vacuum-associated reweighting method, using the gauge configurations around each vacuum separately. Finally, we obtain the Polyakov-loop effective potential, which is well depicted around the $Z_3$-symmetric and $Z_3$-broken vacua., Comment: 11 pages, 4 figures, prepared for XIIth Int. Conf. on New Frontiers in Physics (ICNFP2023)

Motivated by one-dimensional color-electric flux-tube formation in four-dimensional (4D) QCD, we investigate a possibility of effective dimensional reduction in the 4D Yang-Mills (YM) theory. We propose a new gauge fixing of "dimensional reduction (DR) gauge" defined so as to minimize $R_{\mathrm{DR}}~\equiv~\int d^{4}s ~ \mathrm{Tr} \left[ A_{x}^{2}(s) + A_{y}^{2}(s) \right]$, which has a residual gauge symmetry for the gauge function $\Omega (t,z)$ like 2D QCD on the $t$-$z$ plane. We investigate effective dimensional reduction in the DR gauge using SU(3) quenched lattice QCD at $\beta = 6.0$. The amplitude of $A_{x}(s)$ and $A_{y}(s)$ are found to be strongly suppressed in the DR gauge. We consider "$tz$-projection" of $A_{x,y}(s) \to 0$ for the gauge configuration generated in the DR gauge, in a similar sense to Abelian projection in the maximally Abelian gauge. By the $tz$-projection in the DR gauge, the interquark potential is not changed, and $A_{t}(s)$ and $A_{z}(s)$ play a dominant role in quark confinement. In the DR gauge, we calculate a spatial correlation $\langle \mathrm{Tr} A_{\perp}(s) A_{\perp}(s+ra_{\perp}) \rangle ~ (\perp = x,y)$ and estimate the spatial mass of $A_{\perp}(s) ~ (\perp = x,y)$ as $M \simeq 1.7 ~ \mathrm{GeV}$. It is conjectured that this large mass makes $A_{\perp}(s)$ inactive and realizes the dominance of $A_{t}(s)$ and $A_{z}(s)$ in infrared region in the DR gauge. We also calculate the spatial correlation of two temporal link-variables and find that the correlation decreases as $\exp (-mr)$ with $m \simeq 0.6 ~ \mathrm{GeV}$. Using a crude approximation, the 4D YM theory is reduced into an ensemble of 2D YM systems with the coupling of $g_{\rm 2D} = g m$., Comment: 15 pages, 11 figures

Neutrinos from very nearby supernovae, such as Betelgeuse, are expected to generate more than ten million events over 10\,s in Super-Kamokande (SK). At such large event rates, the buffers of the SK analog-to-digital conversion board (QBEE) will overflow, causing random loss of data that is critical for understanding the dynamics of the supernova explosion mechanism. In order to solve this problem, two new DAQ modules were developed to aid in the observation of very nearby supernovae. The first of these, the SN module, is designed to save only the number of hit PMTs during a supernova burst and the second, the Veto module, prescales the high rate neutrino events to prevent the QBEE from overflowing based on information from the SN module. In the event of a very nearby supernova, these modules allow SK to reconstruct the time evolution of the neutrino event rate from beginning to end using both QBEE and SN module data. This paper presents the development and testing of these modules together with an analysis of supernova-like data generated with a flashing laser diode. We demonstrate that the Veto module successfully prevents DAQ overflows for Betelgeuse-like supernovae as well as the long-term stability of the new modules. During normal running the Veto module is found to issue DAQ vetos a few times per month resulting in a total dead time less than 1\,ms, and does not influence ordinary operations. Additionally, using simulation data we find that supernovae closer than 800~pc will trigger Veto module resulting in a prescaling of the observed neutrino data., Comment: 28 pages, 18 figures. Submitted to PTEP

We present the results of the charge ratio ($R$) and polarization ($P^{\mu}_{0}$) measurements using the decay electron events collected from 2008 September to 2022 June by the Super-Kamiokande detector. Because of its underground location and long operation, we performed high precision measurements by accumulating cosmic-ray muons. We measured the muon charge ratio to be $R=1.32 \pm 0.02$ $(\mathrm{stat.}{+}\mathrm{syst.})$ at $E_{\mu}\cos \theta_{\mathrm{Zenith}}=0.7^{+0.3}_{-0.2}$ $\mathrm{TeV}$, where $E_{\mu}$ is the muon energy and $\theta_{\mathrm{Zenith}}$ is the zenith angle of incoming cosmic-ray muons. This result is consistent with the Honda flux model while this suggests a tension with the $\pi K$ model of $1.9\sigma$. We also measured the muon polarization at the production location to be $P^{\mu}_{0}=0.52 \pm 0.02$ $(\mathrm{stat.}{+}\mathrm{syst.})$ at the muon momentum of $0.9^{+0.6}_{-0.1}$ $\mathrm{TeV}/c$ at the surface of the mountain; this also suggests a tension with the Honda flux model of $1.5\sigma$. This is the most precise measurement ever to experimentally determine the cosmic-ray muon polarization near $1~\mathrm{TeV}/c$. These measurement results are useful to improve the atmospheric neutrino simulations., Comment: 29 pages, 45 figures

An analysis of solar neutrino data from the fourth phase of Super-Kamiokande~(SK-IV) from October 2008 to May 2018 is performed and the results are presented. The observation time of the data set of SK-IV corresponds to $2970$~days and the total live time for all four phases is $5805$~days. For more precise solar neutrino measurements, several improvements are applied in this analysis: lowering the data acquisition threshold in May 2015, further reduction of the spallation background using neutron clustering events, precise energy reconstruction considering the time variation of the PMT gain. The observed number of solar neutrino events in $3.49$--$19.49$ MeV electron kinetic energy region during SK-IV is $65,443^{+390}_{-388}\,(\mathrm{stat.})\pm 925\,(\mathrm{syst.})$ events. Corresponding $\mathrm{^{8}B}$ solar neutrino flux is $(2.314 \pm 0.014\, \rm{(stat.)} \pm 0.040 \, \rm{(syst.)}) \times 10^{6}~\mathrm{cm^{-2}\,s^{-1}}$, assuming a pure electron-neutrino flavor component without neutrino oscillations. The flux combined with all SK phases up to SK-IV is $(2.336 \pm 0.011\, \rm{(stat.)} \pm 0.043 \, \rm{(syst.)}) \times 10^{6}~\mathrm{cm^{-2}\,s^{-1}}$. Based on the neutrino oscillation analysis from all solar experiments, including the SK $5805$~days data set, the best-fit neutrino oscillation parameters are $\rm{sin^{2} \theta_{12,\,solar}} = 0.306 \pm 0.013 $ and $\Delta m^{2}_{21,\,\mathrm{solar}} = (6.10^{+ 0.95}_{-0.81}) \times 10^{-5}~\rm{eV}^{2}$, with a deviation of about 1.5$\sigma$ from the $\Delta m^{2}_{21}$ parameter obtained by KamLAND. The best-fit neutrino oscillation parameters obtained from all solar experiments and KamLAND are $\sin^{2} \theta_{12,\,\mathrm{global}} = 0.307 \pm 0.012 $ and $\Delta m^{2}_{21,\,\mathrm{global}} = (7.50^{+ 0.19}_{-0.18}) \times 10^{-5}~\rm{eV}^{2}$., Comment: 47 pages, 61 figures

The deconfinement transition in non-Abelian gauge theory is understood as spontaneous breaking of $\mathbb{Z}_N$ symmetry at high temperatures. Accordingly, quark-gluon plasma generally includes some partial cells called center domains, each with a homogeneous Polyakov-loop expectation value. In this work, constructing an effective action describing the deconfinement vacuum of Yang-Mills theory with $N$ colors, we discuss the properties of center domains. First, we evaluate the spatial correlation of local Polyakov-loop fluctuation and demonstrate that some fluctuation becomes a Nambu-Goldstone-like mode in the large-$N$ limit. We also discuss surface tension between two $\mathbb{Z}_N$ center domains. Second, we estimate the global vacuum-to-vacuum transition in a single center domain. We find that some threshold volume exists, where a domain larger than this volume is stable, and vice versa. Identifying the threshold as the lower bound of a stable center domain volume, we quantitatively argue the typical volume scale of center domains., Comment: 7 pages, 4 figures; Proceedings of the 40th International Symposium on Lattice Field Theory (LATTICE2023)

Computer vision has become increasingly prevalent in solving real-world problems across diverse domains, including smart agriculture, fishery, and livestock management. These applications may not require processing many image frames per second, leading practitioners to use single board computers (SBCs). Although many lightweight networks have been developed for mobile/edge devices, they primarily target smartphones with more powerful processors and not SBCs with the low-end CPUs. This paper introduces a CNN-ViT hybrid network called SBCFormer, which achieves high accuracy and fast computation on such low-end CPUs. The hardware constraints of these CPUs make the Transformer's attention mechanism preferable to convolution. However, using attention on low-end CPUs presents a challenge: high-resolution internal feature maps demand excessive computational resources, but reducing their resolution results in the loss of local image details. SBCFormer introduces an architectural design to address this issue. As a result, SBCFormer achieves the highest trade-off between accuracy and speed on a Raspberry Pi 4 Model B with an ARM-Cortex A72 CPU. For the first time, it achieves an ImageNet-1K top-1 accuracy of around 80% at a speed of 1.0 frame/sec on the SBC. Code is available at https://github.com/xyongLu/SBCFormer., Comment: 11 pages, 2 figures, WACV2024

This paper addresses the problem of predicting hazards that drivers may encounter while driving a car. We formulate it as a task of anticipating impending accidents using a single input image captured by car dashcams. Unlike existing approaches to driving hazard prediction that rely on computational simulations or anomaly detection from videos, this study focuses on high-level inference from static images. The problem needs predicting and reasoning about future events based on uncertain observations, which falls under visual abductive reasoning. To enable research in this understudied area, a new dataset named the DHPR (Driving Hazard Prediction and Reasoning) dataset is created. The dataset consists of 15K dashcam images of street scenes, and each image is associated with a tuple containing car speed, a hypothesized hazard description, and visual entities present in the scene. These are annotated by human annotators, who identify risky scenes and provide descriptions of potential accidents that could occur a few seconds later. We present several baseline methods and evaluate their performance on our dataset, identifying remaining issues and discussing future directions. This study contributes to the field by introducing a novel problem formulation and dataset, enabling researchers to explore the potential of multi-modal AI for driving hazard prediction., Comment: Main Paper: 11 pages, Supplementary Materials: 25 pages

We investigate a baryon and its dilatation modes in holographic QCD based on the Sakai-Sugimoto model, which is expressed as a 1+4 dimensional U($N_f$) gauge theory in the flavor space. For spatially rotational symmetric systems, we apply a generalized version of the Witten Ansatz, and reduce 1+4 dimensional holographic QCD into a 1+2 dimensional Abelian Higgs theory in a curved space. In the reduced theory, the holographic baryon is described as a two-dimensional topological object of an Abrikosov vortex. We numerically calculate the baryon solution of holographic QCD using a fine and large lattice with spacing of 0.04 fm and size of 10 fm. Using the relation between the baryon size and the zero-point location of the Higgs field in the description with the Witten Ansatz, we investigate a various-size baryon through this vortex description. As time-dependent size-oscillation modes (dilatation modes) of a baryon, we numerically obtain the lowest excitation energy of 577 MeV and deduce the dilatational excitation of a nucleon to be the Roper resonance N$^*$(1440)., Comment: 22 pages, 14 figures

Recent studies on visual anomaly detection (AD) of industrial objects/textures have achieved quite good performance. They consider an unsupervised setting, specifically the one-class setting, in which we assume the availability of a set of normal (\textit{i.e.}, anomaly-free) images for training. In this paper, we consider a more challenging scenario of unsupervised AD, in which we detect anomalies in a given set of images that might contain both normal and anomalous samples. The setting does not assume the availability of known normal data and thus is completely free from human annotation, which differs from the standard AD considered in recent studies. For clarity, we call the setting blind anomaly detection (BAD). We show that BAD can be converted into a local outlier detection problem and propose a novel method named PatchCluster that can accurately detect image- and pixel-level anomalies. Experimental results show that PatchCluster shows a promising performance without the knowledge of normal data, even comparable to the SOTA methods applied in the one-class setting needing it.

Previous works on unsupervised industrial anomaly detection mainly focus on local structural anomalies such as cracks and color contamination. While achieving significantly high detection performance on this kind of anomaly, they are faced with logical anomalies that violate the long-range dependencies such as a normal object placed in the wrong position. In this paper, based on previous knowledge distillation works, we propose to use two students (local and global) to better mimic the teacher's behavior. The local student, which is used in previous studies mainly focuses on structural anomaly detection while the global student pays attention to logical anomalies. To further encourage the global student's learning to capture long-range dependencies, we design the global context condensing block (GCCB) and propose a contextual affinity loss for the student training and anomaly scoring. Experimental results show the proposed method doesn't need cumbersome training techniques and achieves a new state-of-the-art performance on the MVTec LOCO AD dataset.

Smartphones equipped with a multi-camera system comprising multiple cameras with different field-of-view (FoVs) are becoming more prevalent. These camera configurations are compatible with reference-based SR and video SR, which can be executed simultaneously while recording video on the device. Thus, combining these two SR methods can improve image quality. Recently, Lee et al. have presented such a method, RefVSR. In this paper, we consider how to optimally utilize the observations obtained, including input low-resolution (LR) video and reference (Ref) video. RefVSR extends conventional video SR quite simply, aggregating the LR and Ref inputs over time in a single bidirectional stream. However, considering the content difference between LR and Ref images due to their FoVs, we can derive the maximum information from the two image sequences by aggregating them independently in the temporal direction. Then, we propose an improved method, RefVSR++, which can aggregate two features in parallel in the temporal direction, one for aggregating the fused LR and Ref inputs and the other for Ref inputs over time. Furthermore, we equip RefVSR++ with enhanced mechanisms to align image features over time, which is the key to the success of video SR. We experimentally show that RefVSR++ outperforms RefVSR by over 1dB in PSNR, achieving the new state-of-the-art.

Despite the recent advancement in the study of removing motion blur in an image, it is still hard to deal with strong blurs. While there are limits in removing blurs from a single image, it has more potential to use multiple images, e.g., using an additional image as a reference to deblur a blurry image. A typical setting is deburring an image using a nearby sharp image(s) in a video sequence, as in the studies of video deblurring. This paper proposes a better method to use the information present in a reference image. The method does not need a strong assumption on the reference image. We can utilize an alternative shot of the identical scene, just like in video deblurring, or we can even employ a distinct image from another scene. Our method first matches local patches of the target and reference images and then fuses their features to estimate a sharp image. We employ a patch-based feature matching strategy to solve the difficult problem of matching the blurry image with the sharp reference. Our method can be integrated into pre-existing networks designed for single image deblurring. The experimental results show the effectiveness of the proposed method.

We report the first search result for the flux of astrophysical electron antineutrinos for energies O(10) MeV in the gadolinium-loaded Super-Kamiokande (SK) detector. In June 2020, gadolinium was introduced to the ultra-pure water of the SK detector in order to detect neutrons more efficiently. In this new experimental phase, SK-Gd, we can search for electron antineutrinos via inverse beta decay with efficient background rejection and higher signal efficiency thanks to the high efficiency of the neutron tagging technique. In this paper, we report the result for the initial stage of SK-Gd with a $22.5\times552$ $\rm kton\cdot day$ exposure at 0.01% Gd mass concentration. No significant excess over the expected background in the observed events is found for the neutrino energies below 31.3 MeV. Thus, the flux upper limits are placed at the 90% confidence level. The limits and sensitivities are already comparable with the previous SK result with pure-water ($22.5 \times 2970 \rm kton\cdot day$) owing to the enhanced neutron tagging.

Using the Polyakov-loop effective action, we investigate the structure of spontaneously broken $\mathbb{Z}_N$ symmetry in the deconfinement vacuum in the SU($N$) Yang-Mills theory with finite $N$. First, we examine the Polyakov-loop fluctuation around a $\mathbb{Z}_N$-broken vacuum and calculate the spatial correlation of its phase variable. We show that the phase variable of the Polyakov loop becomes a Nambu-Goldstone mode in the large-$N$ limit. Second, we estimate the global vacuum-to-vacuum transition rate in a finite-volume domain of the quark-gluon plasma. Based on our estimation, we state that some threshold volume exists, a domain larger than which is stable, and vice versa. Identifying the threshold as the lower bound of a stable center domain volume, we find the typical volume scale of center domains., Comment: 14 pages, 9 figures

Using SU(3) lattice QCD, we investigate role of spatial gluons for hadron masses in the Coulomb gauge, considering the relation between QCD and the quark model. From the Coulomb-gauge configurations at the quenched level on a $16^3 \times 32$ lattice at $\beta$ = 6.0, we consider the $\vec{A} = 0$ projection, where all the spatial gluon fields are set to zero. In this projection, the inter-quark potential is unchanged. We investigate light hadron masses and find that nucleon and delta baryon masses are almost degenerate. This result suggests that the N-$\Delta$ mass difference arises from the color-magnetic interactions, which is consistent with the quark model picture. Next, as a generalization of this projection, we expand spatial gluon fields in terms of Faddeev-Popov eigenmodes and leave only some partial components. We find that the ${\rm N}-\Delta$ and $0^{++}-2^{++}$ glueball mass splittings are almost reproduced only with 1 \% low-lying components. This suggests that low-lying color-magnetic interaction leads to the hadron mass splitting., Comment: 8 pages, 5 figures, contribution to the proceedings of the 15th Quark Confinement and the Hadron Spectrum conference (ConfXV), 1st-6th August 2022, Stavanger, Norway

Cosmic-ray muons that enter the Super-Kamiokande detector cause hadronic showers due to spallation in water, producing neutrons and radioactive isotopes. Those are a major background source for studies of MeV-scale neutrinos and searches for rare events. Since 2020, gadolinium was introduced in the ultra-pure water in the Super-Kamiokande detector to improve the detection efficiency of neutrons. In this study, the cosmogenic neutron yield was measured using data acquired during the period after the gadolinium loading. The yield was found to be $(2.76 \pm 0.02\,\mathrm{(stat.) \pm 0.19\,\mathrm{(syst.)}}) \times 10^{-4}\,\mu^{-1} \mathrm{g^{-1} cm^{2}}$ at 259 GeV of average muon energy at the Super-Kamiokande detector., Comment: 10 pages, 10 figures, 3 tables

Neutrinos associated with solar flares (solar-flare neutrinos) provide information on particle acceleration mechanisms during the impulsive phase of solar flares. We searched using the Super-Kamiokande detector for neutrinos from solar flares that occurred during solar cycles $23$ and $24$, including the largest solar flare (X28.0) on November 4th, 2003. In order to minimize the background rate we searched for neutrino interactions within narrow time windows coincident with $\gamma$-rays and soft X-rays recorded by satellites. In addition, we performed the first attempt to search for solar-flare neutrinos from solar flares on the invisible side of the Sun by using the emission time of coronal mass ejections (CMEs). By selecting twenty powerful solar flares above X5.0 on the visible side and eight CMEs whose emission speed exceeds $2000$ $\mathrm{km \, s^{-1}}$ on the invisible side from 1996 to 2018, we found two (six) neutrino events coincident with solar flares occurring on the visible (invisible) side of the Sun, with a typical background rate of $0.10$ ($0.62$) events per flare in the MeV-GeV energy range. No significant solar-flare neutrino signal above the estimated background rate was observed. As a result we set the following upper limit on neutrino fluence at the Earth $\mathit{\Phi}<1.1\times10^{6}$ $\mathrm{cm^{-2}}$ at the $90\%$ confidence level for the largest solar flare. The resulting fluence limits allow us to constrain some of the theoretical models for solar-flare neutrino emission., Comment: 36 pages, 18 figures, 9 tables (Figure 12 was replaced because it was incorrect in version 1.)

We report a search for cosmic-ray boosted dark matter with protons using the 0.37 megaton$\times$years data collected at Super-Kamiokande experiment during the 1996-2018 period (SKI-IV phase). We searched for an excess of proton recoils above the atmospheric neutrino background from the vicinity of the Galactic Center. No such excess is observed, and limits are calculated for two reference models of dark matter with either a constant interaction cross-section or through a scalar mediator. This is the first experimental search for boosted dark matter with hadrons using directional information. The results present the most stringent limits on cosmic-ray boosted dark matter and exclude the dark matter-nucleon elastic scattering cross-section between $10^{-33}\text{ cm}^{2}$ and $10^{-27}\text{ cm}^{2}$ for dark matter mass from 10 MeV/$c^2$ to 1 GeV/$c^2$., Comment: With 1-page appendix. A bug was found in July 2023. This version is updated to match the erratum

In this paper, we present a new neural architectural block for the vision domain, named Mixing Regionally and Locally (MRL), developed with the aim of effectively and efficiently mixing the provided input features. We bifurcate the input feature mixing task as mixing at a regional and local scale. To achieve an efficient mix, we exploit the domain-wide receptive field provided by self-attention for regional-scale mixing and convolutional kernels restricted to local scale for local-scale mixing. More specifically, our proposed method mixes regional features associated with local features within a defined region, followed by a local-scale features mix augmented by regional features. Experiments show that this hybridization of self-attention and convolution brings improved capacity, generalization (right inductive bias), and efficiency. Under similar network settings, MRL outperforms or is at par with its counterparts in classification, object detection, and segmentation tasks. We also show that our MRL-based network architecture achieves state-of-the-art performance for H&E histology datasets. We achieved DICE of 0.843, 0.855, and 0.892 for Kumar, CoNSep, and CPM-17 datasets, respectively, while highlighting the versatility offered by the MRL framework by incorporating layers like group convolutions to improve dataset-specific generalization.

We searched for proton decay via $p\to\mu^+K^0$ in 0.37\,Mton$\cdot$years of data collected between 1996 and 2018 from the Super-Kamiokande water Cherenkov experiment. The selection criteria were defined separately for $K^0_S$ and $K^0_L$ channels. No significant event excess has been observed. As a result of this analysis, which extends the previous search by an additional 0.2\,Mton$\cdot$years of exposure and uses an improved event reconstruction, we set a lower limit of $3.6\times10^{33}$ years on the proton lifetime., Comment: 13 pages, 11 figures

Aiming at the relation between QCD and the quark model, we consider projections of gauge configurations generated in quenched lattice QCD simulations in the Coulomb gauge on a 16$^{\rm 3}$ $\rm \times$ 32, $\rm \beta$ = 6.0 lattice. First, we focus on a fact that the static quark-antiquark potential is independent of spatial gauge fields. We explicitly confirm this by performing $\vec{A}$ = 0 projection, where spatial gauge fields are all set to zero. We also apply the $\vec{A}$ = 0 projection to light hadron masses and find that nucleon and delta baryon masses are almost degenerate, suggesting vanishing of the color-magnetic interactions. After considering the physical meaning of the $\vec{A}$ = 0 projection, we next propose a generalized projection, where spatial gauge fields are expanded in terms of Faddeev-Popov eigenmodes and only some eigenmodes are left. We apply the proposed projection to light hadron and glueball masses and find that the N-$\rm \Delta$ and 0$^{\rm ++}$-2$^{\rm ++}$ mass splittings become evident when projected with more than 33 (0.10 \%) low-lying eigenmodes, suggesting emergence of the color-magnetic interactions. We also find that the original hadron masses are approximately reproduced with just 328 (1.00 \%) low-lying eigenmodes. These findings indicate an important role of low-lying eigenmodes on hadron masses and would be useful in clarifying the relation between QCD and the quark model., Comment: 22 pages, 9 figures, and 2 tables, to appear in EPJA

Open-set object detection (OSOD) has recently gained attention. It is to detect unknown objects while correctly detecting known objects. In this paper, we first point out that the recent studies' formalization of OSOD, which generalizes open-set recognition (OSR) and thus considers an unlimited variety of unknown objects, has a fundamental issue. This issue emerges from the difference between image classification and object detection, making it hard to evaluate OSOD methods' performance properly. We then introduce a novel scenario of OSOD, which considers known and unknown classes within a specified super-class of object classes. This new scenario has practical applications and is free from the above issue, enabling proper evaluation of OSOD performance and probably making the problem more manageable. Finally, we experimentally evaluate existing OSOD methods with the new scenario using multiple datasets, showing that the current state-of-the-art OSOD methods attain limited performance similar to a simple baseline method. The paper also presents a taxonomy of OSOD that clarifies different problem formalizations. We hope our study helps the community reconsider OSOD problems and progress in the right direction., Comment: 17 pages, 7 figures

Current state-of-the-art methods for image captioning employ region-based features, as they provide object-level information that is essential to describe the content of images; they are usually extracted by an object detector such as Faster R-CNN. However, they have several issues, such as lack of contextual information, the risk of inaccurate detection, and the high computational cost. The first two could be resolved by additionally using grid-based features. However, how to extract and fuse these two types of features is uncharted. This paper proposes a Transformer-only neural architecture, dubbed GRIT (Grid- and Region-based Image captioning Transformer), that effectively utilizes the two visual features to generate better captions. GRIT replaces the CNN-based detector employed in previous methods with a DETR-based one, making it computationally faster. Moreover, its monolithic design consisting only of Transformers enables end-to-end training of the model. This innovative design and the integration of the dual visual features bring about significant performance improvement. The experimental results on several image captioning benchmarks show that GRIT outperforms previous methods in inference accuracy and speed., Comment: Accepted to ECCV 2022; 14 pages with appendix; Code: https://github.com/davidnvq/grit

Reconstructing ghosting-free high dynamic range (HDR) images of dynamic scenes from a set of multi-exposure images is a challenging task, especially with large object motion and occlusions, leading to visible artifacts using existing methods. To address this problem, we propose a deep network that tries to learn multi-scale feature flow guided by the regularized loss. It first extracts multi-scale features and then aligns features from non-reference images. After alignment, we use residual channel attention blocks to merge the features from different images. Extensive qualitative and quantitative comparisons show that our approach achieves state-of-the-art performance and produces excellent results where color artifacts and geometric distortions are significantly reduced.

In this paper, we consider the problem in defocus image deblurring. Previous classical methods follow two-steps approaches, i.e., first defocus map estimation and then the non-blind deblurring. In the era of deep learning, some researchers have tried to address these two problems by CNN. However, the simple concatenation of defocus map, which represents the blur level, leads to suboptimal performance. Considering the spatial variant property of the defocus blur and the blur level indicated in the defocus map, we employ the defocus map as conditional guidance to adjust the features from the input blurring images instead of simple concatenation. Then we propose a simple but effective network with spatial modulation based on the defocus map. To achieve this, we design a network consisting of three sub-networks, including the defocus map estimation network, a condition network that encodes the defocus map into condition features, and the defocus deblurring network that performs spatially dynamic modulation based on the condition features. Moreover, the spatially dynamic modulation is based on an affine transform function to adjust the features from the input blurry images. Experimental results show that our method can achieve better quantitative and qualitative evaluation performance than the existing state-of-the-art methods on the commonly used public test datasets.

Unsupervised domain adaptation (UDA) adapts a model trained on one domain (called source) to a novel domain (called target) using only unlabeled data. Due to its high annotation cost, researchers have developed many UDA methods for semantic segmentation, which assume no labeled sample is available in the target domain. We question the practicality of this assumption for two reasons. First, after training a model with a UDA method, we must somehow verify the model before deployment. Second, UDA methods have at least a few hyper-parameters that need to be determined. The surest solution to these is to evaluate the model using validation data, i.e., a certain amount of labeled target-domain samples. This question about the basic assumption of UDA leads us to rethink UDA from a data-centric point of view. Specifically, we assume we have access to a minimum level of labeled data. Then, we ask how much is necessary to find good hyper-parameters of existing UDA methods. We then consider what if we use the same data for supervised training of the same model, e.g., finetuning. We conducted experiments to answer these questions with popular scenarios, {GTA5, SYNTHIA}$\rightarrow$Cityscapes. We found that i) choosing good hyper-parameters needs only a few labeled images for some UDA methods whereas a lot more for others; and ii) simple finetuning works surprisingly well; it outperforms many UDA methods if only several dozens of labeled images are available., Comment: Under review in Pattern Recognition Letters

Super-Kamiokande has been searching for neutrino bursts characteristic of core-collapse supernovae continuously, in real time, since the start of operations in 1996. The present work focuses on detecting more distant supernovae whose event rate may be too small to trigger in real time, but may be identified using an offline approach. The analysis of data collected from 2008 to 2018 found no evidence of distant supernovae bursts. This establishes an upper limit of 0.29 year$^{-1}$ on the rate of core-collapse supernovae out to 100 kpc at 90% C.L.. For supernovae that fail to explode and collapse directly to black holes the limit reaches to 300 kpc.