Kernel Shape Research Articles

Nonlinear Meshfree Collocation Method for Solving Double-Diffusive Natural Convection in a Porous Enclosure

A meshfree collocation method formulated by using reproducing kernel shape functions with Newton–Raphson method is presented with application to analyzing a three-phase coupling system in a porous enclosure. To tackle such a highly nonlinear phenomenon caused by double-diffusive natural convection, direct collocation using local approximation in conjunction with a two-step iteration solver is employed to stably and efficiently solve this problem. It has been shown that a determined and well-conditioned system is established under the collocation framework with reproducing kernel approximation. By conducting the parametric study, the numerical results have demonstrated the robustness of the proposed method in analyzing the porous enclosure by considering various geometry, boundary conditions, and governing parameters.

Dynamic convolutional time series forecasting based on adaptive temporal bilateral filtering

Time series data typically contain complex dynamic patterns, which not only include linear trends and seasonal variations but also significant nonlinear changes and complex dependencies. Currently, feature extraction methods for time series data primarily employ mixed-mode extraction within the temporal domain, neglecting the effective extraction and analysis of nonlinear characteristics between different observation points and the interdependencies between variables. To address these issues, we designed an adaptive temporal bilateral filtering module that effectively preserves and highlights the nonlinear features and patterns in time series while filtering out noise and redundant information. We also designed a nonlinear feature adaptive extraction module that integrates a gating mechanism and deformable convolutions. This design allows the model to adaptively adjust the shape of the convolutional kernels according to different time steps and conditions. This enables accurate capture and extraction of nonlinear features between various time observations and dependencies between different variables. Additionally, we use stacked convolutional layers to extract local contextual features, addressing fluctuations in local features caused by changes in data distribution in real-world scenarios. In summary, we propose DCNet, a dynamic convolutional network based on an adaptive temporal bilateral filter, and evaluated its performance on twelve real-world datasets. The results indicate that DCNet consistently achieves state-of-the-art performance in both short-term and long-term forecasting tasks, with favorable runtime efficiency.

Interplay of kernel shape and surface structure for NIR luminescence in atomically precise gold nanorods

It is challenging to attain strong near-infrared (NIR) emissive gold nanoclusters. Here we show a rod-shaped cluster with the composition of [Au28(p-MBT)14(Hdppa)3](SO3CF3)2 (1 for short, Hdppa is N,N-bis(diphenylphosphino)amine, p-MBT is 4-methylbenzenethiolate) has been synthesized. Single crystal X-ray structural analysis reveals that it has a rod-like face-centered cubic (fcc) Au22 kernel built from two interpenetrating bicapped cuboctahedral Au15 units. 1 features NIR luminescence with an emission maximum at 920 nm, and the photoluminescence quantum yield (PLQY) is 12%, which is 30-fold of [Au21(m-MBT)12(Hdppa)2]SO3CF3 (2, m-MBT is 3-methylbenzenethiolate) with a similar composition and 60-fold of Au30S(S‑t‑Bu)18 with a similar structure. time-dependent DFT(TDDFT)calculations reveal that the luminescence of 1 is associated with the Au22 kernel. The small Stokes shift of 1 indicates that it has a very small excited state structural distortion, leading to high radiative decay rate (kr) probability. The emission of cluster 1 is a mixture of phosphorescence and thermally activated delayed fluorescence(TADF), and the enhancement of the NIR emission is mainly due to the promotion of kr rather than the inhibition of knr. This work demonstrates that the metal kernel and the surface structure are both very important for cluster-based NIR luminescence materials.

Building detection based on searching of the optimal kernel shapes pruning method on Res2-Unet

<p>In recent years, advances in remote sensing technology have made it feasible to use satellite data for large-scale building detection. Moreover, the building detection from multispectral satellite photography data is necessary; however, it is difficult to recovery the accurate building footprint from the high-resolution pictures. Because the deep learning networks contains high computational cost and over-parameterized. Therefore, network pruning has been used to reduce the storage and computations of convolutional neural network (CNN) models. In this article, we proposed the pruning technique to prune the CNN network from Res2-Unet model for accurately detecting the buildings. Initially, the CNN network is pruned by utilizing the searching of the optimal kernel shapes technique. It is employed to carry out stripe-wise pruning and automatically find the ideal kernel shapes. Then the data quantification is applied to enhance the proposed model and also reduce the complexity. Finally, the enhanced Res2-Unet model is used for the building detection. Moreover, WHU East Asia Satellite and the Massachusetts building dataset are the two available datasets used to access the suggested framework. Compare to the existing models, the proposed model gives better performance.</p>

Propagation of Expanding Ellipsoid-Shaped Flame

ABSTRACT Laminar flame speed is commonly measured using the spherically expanding flame technique. The non-ideal shape of the initial flame kernel may have a significant influence on the subsequent flame propagation. In the present study, the propagation evolution of expanding ellipsoidal flame initiated from an ellipsoidal flame kernel was simulated with the propagation speed following the linear stretch relation. The effects of two components for the total stretch rate, i.e. curvature stretch and tangential stretch, were investigated. Different from the ideal spherical flame, which excludes the effect of tangential flow, tangential stretch can be negative and can be on the same order as the curvature stretch during the early stage of flame propagation. With flame expanding outwardly, the absolute value of tangential stretch decreases. The relative decrement of tangential stretch is more obvious if the flame speed is faster. The present results imply that large initial flame radius data should be used for laminar flame speed extrapolation when the shape of the flame significantly deviates from an ideal sphere.

Kernel Trait and Genetic Variability Analysis in BC2F1 Maize Lines Derived from Zea nicaraguensis Crosses

Maize (Zea mays L.) is an important cereal crop with significant applications in food, feed and industrial applications. This study aimed to increase maize yield and quality by evaluating core traits in introgressed Zea nicaraguensis individual’s recipients of LM13 and LM14 maize inbred lines. Kernel type, shape, color, dimensions and weight were assessed in 35 BC2F1 plants using standardized descriptor guidelines. Analysis of genotypic and phenotypic coefficients of variation, heritability and genetic progress showed significant genetic diversity and the influence of environmental factors. Significant sources of variability highlighted by principal component analysis were kernel color, endosperm color, and 100-seed weight. High heritability and genetic progress in kernel color and length-to-width ratio indicate great selection potential. Reintroducing the diversity lost in cultivated maize inherent in Zea nicaraguensis has the potential to improve both its agronomic performance and quality. Future research should aim to identify genetic markers and examine the interaction between genetic and environmental influences to optimize maize varieties for specific environments and consumer preferences, ultimately contributing to global food security and agricultural sustainability.

AFE-YOLOv8: A Novel Object Detection Model for Unmanned Aerial Vehicle Scenes with Adaptive Feature Enhancement

Object detection in unmanned aerial vehicle (UAV) scenes is a challenging task due to the varying scales and complexities of targets. To address this, we propose a novel object detection model, AFE-YOLOv8, which integrates three innovative modules: the Multi-scale Nonlinear Fusion Module (MNFM), the Adaptive Feature Enhancement Module (AFEM), and the Receptive Field Expansion Module (RFEM). The MNFM introduces nonlinear mapping by exploiting the property that deformable convolution can dynamically adjust the shape of the convolution kernel according to the shape of the target, and it effectively enhances the feature extraction capability of the backbone network by integrating multi-scale feature maps from different mapping branches. Meanwhile, the AFEM introduces an adaptive fusion factor, and through the fusion factor, it adaptively integrates the small-target features contained in the feature maps of different detection branches into the small-target detection branch, thus enhancing the expression of the small-target features contained in the feature maps of the small-target detection branch. Furthermore, the RFEM expands the receptive field of the feature maps of the large- and medium-scale target detection branches through stacked convolution, so as to make the model’s receptive field cover the whole target, and thereby learn more rich and comprehensive features of the target. The experimental results demonstrate the superior performance of the proposed model compared to the baseline in detecting objects of various scales. On the VisDrone dataset, the proposed model achieves a 4.5% enhancement in mean average precision (mAP) and a 5.45% improvement in average precision at an IOU threshold of 0.5 (AP50). Additionally, ablation experiments conducted on the challenging DOTA dataset showcase the model’s robustness and generalization capabilities.

CSD-YOLO: A Ship Detection Algorithm Based on a Deformable Large Kernel Attention Mechanism

Ship detection and identification play pivotal roles in ensuring navigation safety and facilitating efficient maritime traffic management. Aiming at ship detection in complex environments, which often faces problems such as the dense occlusion of ship targets, low detection accuracy, and variable environmental conditions, in this paper, we propose a ship detection algorithm CSD-YOLO (Context guided block module, Slim-neck, Deformable large kernel attention-You Only Look Once) based on the deformable large kernel attention (D-LKA) mechanism, which was improved based on YOLOv8 to enhance its performance. This approach integrates several innovations to bolster its performance. Initially, the utilization of the Context Guided Block module (CG block) enhanced the c2f module of the backbone network, thereby augmenting the feature extraction capabilities and enabling a more precise capture of the key image information. Subsequently, the introduction of a novel neck architecture and the incorporation of the slim-neck module facilitated more effective feature fusion, thereby enhancing both the accuracy and efficiency of detection. Furthermore, the algorithm incorporates a D-LKA mechanism to dynamically adjust the convolution kernel shape and size, thereby enhancing the model’s adaptability to varying ship target shapes and sizes. To address data scarcity in complex marine environments, the experiments utilized a fused dataset comprising the SeaShips dataset and a proprietary dataset. The experimental results demonstrate that the CSD-YOLO algorithm outperformed the YOLOv8n algorithm across all model evaluation metrics. Specifically, the precision rate (precision) was 91.5%, the recall rate (recall) was 89.5%, and the mean accuracy (mAP) was 91.5%. Compared to the benchmark algorithm, the Recall was improved by 0.7% and the mAP was improved by 0.4%. These results indicate that the CSD-YOLO algorithm can effectively meet the requirements for ship target recognition and tracking in complex marine environments.

Diversity-Learning Block: Conquer Feature Homogenization of Multibranch.

Visual Geometry Group (VGG)-style ConvNet is an neural-network process units (NPU)-friendly network; however, the accuracy of this architecture cannot keep up with other well-designed network structures. Although some reparameterization methods are proposed to remedy this weakness, their performance suffers from the homogenization issue of parallel branches, and the preset shape of convolution kernels also influences spatial perception. To address this problem, we propose a diversity-learning (DL) block to build the DLNet, which could adaptively learn various features to enrich the feature space. To balance floating point of operations (FLOPs) and accuracy, groupwise operation is introduced and finally, a lightweight DL ConvNet DLGNet is obtained. Extensive evaluations have been conducted on different computer vision tasks, e.g., image classification [Canadian Institute For Advanced Research (CIFAR) and ImageNet], object detection [PASCAL visual object classes (VOC) and Microsoft Common Objects in Context (MS COCO)], and semantic segmentation (Cityscapes). The experimental results show that our proposed DLGNet can achieve comparable performance with the state-of-the-art networks while the speed is 183% faster than GhostNet and even over 600% faster than MobileNetV3 with similar accuracy when running on NPU.

Laser-induced indirect ignition of non-premixed turbulent shear layers

This study uses direct numerical simulations to investigate the forced ignition of temporally-evolving turbulent subsonic shear layers separating a gaseous stream of methane (CH4) and a stagnant gas environment of molecular oxygen (O2). Ignition is forced by a thermal-energy source that heats up a small volume of gas during periods of time much shorter than the characteristic acoustic time scale. The kernel, including its initial rounded conical shape, resembles experimental observations after optical breakdown is achieved in a gas irradiated by a focused laser. Particular emphasis is placed on ignition phenomena observed when the laser is focused on the O2 environment outside the turbulent shear layer, where the local composition is far beyond the lean flammability limit. This represents an indirect mode of non-premixed ignition that develops after a relatively long period of time has passed since laser-energy deposition, when the eddies near the oxidizer edge of the turbulent shear layer are intercepted by a baroclinically-generated subsonic vortex of hot dissociated O2 ejected from the laser-energy deposition zone. The success of indirect ignition depends on the orientation of the laser beam, the thickness of the shear layer, and the kernel standoff distance from the oxidizer edge of the shear layer. An ignition Damköhler number is defined that accounts for these averaged effects in six different simulation cases. For near-unity ignition Damköhler numbers, the solution is also sensitive to the local instantaneous flow field prevailing near the oxidizer edge of the shear layer, in that a modification of the instantaneous pre-deposition flow field, while preserving its turbulence statistics, can produce different ignition outcomes.Novelty and significance statement This work investigates forced ignition in a canonical flow configuration – a non-premixed turbulent CH4/O2 shear layer – and analyzes mechanisms by which ignition can be unexpectedly achieved in adverse conditions. In particular, an ensuing laser-generated advection of hot dissociated gas is observed to ignite the shear layer even when the mixture at the laser focal point is not flammable, an observation first made in early experiments. This flow phenomenon, whose consequences on ignition of non-premixed turbulent flows have not previously been analyzed by direct numerical simulation, can significantly separate the laser-energy deposition zone from the ignition location and is linked to baroclinically generated vorticity due to the rounded conical shape of the energy kernel. The phenomenology analyzed here may impact the design of laser-based ignition systems for chemical propulsion.

Deep local-to-global feature learning for medical image super-resolution

Medical images play a vital role in medical analysis by providing crucial information about patients’ pathological conditions. However, the quality of these images can be compromised by many factors, such as limited resolution of the instruments, artifacts caused by movements, and the complexity of the scanned areas. As a result, low-resolution (LR) images cannot provide sufficient information for diagnosis. To address this issue, researchers have attempted to apply image super-resolution (SR) techniques to restore the high-resolution (HR) images from their LR counterparts. However, these techniques are designed for generic images, and thus suffer from many challenges unique to medical images. An obvious one is the diversity of the scanned objects; for example, the organs, tissues, and vessels typically appear in different sizes and shapes, and are thus hard to restore with standard convolution neural networks (CNNs). In this paper, we develop a dynamic-local learning framework to capture the details of these diverse areas, consisting of deformable convolutions with adjustable kernel shapes. Moreover, the global information between the tissues and organs is vital for medical diagnosis. To preserve global information, we propose pixel–pixel and patch–patch global learning using a non-local mechanism and a vision transformer (ViT), respectively. The result is a novel CNN-ViT neural network with Local-to-Global feature learning for medical image SR, referred to as LGSR, which can accurately restore both local details and global information. We evaluate our method on six public datasets and one large-scale private dataset, which include five different types of medical images (i.e., Ultrasound, OCT, Endoscope, CT, and MRI images). Experiments show that the proposed method achieves superior PSNR/SSIM and visual performance than the state of the arts with competitive computational costs, measured in network parameters, runtime, and FLOPs. What is more, the experiment conducted on OCT image segmentation for the downstream task demonstrates a significantly positive performance effect of LGSR.

Fine mapping of a major QTL, qKl-1BL controlling kernel length in common wheat.

A major stable QTL, qKl-1BL, for kernel length of wheat was narrowed down to a 2.04-Mb interval on chromosome 1BL; the candidate genes were predicated and the genetic effects on yield-related traits were characterized. As a key factor influencing kernel weight, wheat kernel shape is closely related to yield formation, and in turn affects both wheat processing quality and market value. Fine mapping of the major quantitative trait loci (QTL) for kernel shape could provide genetic resources and a theoretical basis for the genetic improvement of wheat yield-related traits. In this study, a major QTL for kernel length (KL) on 1BL, named qKl-1BL, was identified from the recombinant inbred lines (RIL) in multiple environments based on the genetic map and physical map, with 4.76-21.15% of the phenotypic variation explained. To fine map qKl-1BL, the map-based cloning strategy was used. By using developed InDel markers, the near-isogenic line (NIL) pairs and eight key recombinants were identified from a segregating population containing 3621 individualsderived from residual heterozygous lines (RHLs)self-crossing. In combination with phenotype identification, qKl-1BL was finely positioned into a 2.04-Mb interval, KN1B:698.15-700.19Mb, with eight differentially expressed genes enriched at the key period of kernel elongation. Based on transcriptome analysis and functional annotation information, two candidate genes for qKl-1BL controlling kernel elongation were identified. Additionally, genetic effect analysis showed that the superior allele of qKl-1BL from Jing411 could increase KL, thousand kernel weight (TKW), and yield per plant (YPP) significantly, as well as kernel bulk density and stability time. Taken together, this study identified a QTL interval for controlling kernel length with two possible candidate genes, whichprovides an important basis for qKl-1BL cloning, functional analysis, and application in molecular breeding programs.

Restoration of vibration-induced remote sensing images based on CNN identification: a comparative approach

ABSTRACT Satellite vibrations during the imaging process frequently cause images to blur. Satellite micro-vibration types vary according to their sources and each type causes a certain form of image blur. In this paper, the effectiveness of using Convolution neural networks (CNNs) in identifying the types of micro-vibration that lead to image blur is addressed. Additionally, a comparative study between three restoration techniques for the restoration of vibration-induced remote sensing images is considered. The utilised restoration technique is based on a two-steps approach: the first step is the identification of the type of vibration using CNNs to identify the kernel shape or the point spread function that causes image degradation. Hence, the kernel function that blurs the image is accurately determined. The second step is the restoration of the degraded image using three different techniques including Modified Levenberg – Marquardt (MLM), Approximate Land-weber (AL), and Total Variation (TV). The restored images were evaluated using four image quality metrics. MLM technique is designed to improve the convergence and accuracy of the deblurring process. The key finding of this paper features the effectiveness of the MLM technique in image restoration compared to AL and TV techniques.

Research on laser induced plasma ignition of gas oxygen/methane

Laser-induced plasma ignition is a new ignition method to achieve stable and reliable ignition of liquid rocket engines. It has the advantages of accurate and controllable ignition position, broadening ignition lean-burn limit and small disturbance to flow field. In this paper, the laser ignition process and combustion characteristics of gas oxygen/gas methane under rocket engine conditions are studied from four aspects, namely different ignition positions, ignition energy, fuel-oxygen momentum ratio and chamber pressure. Using a high-speed camera, the dynamic process of ignition was captured by the shadow method, and the generation and propagation of laser-induced flame nuclei were recorded. The results show that at the airflow velocity of 300–400 m/s, the time required for the flame kernel to develop to the entire combustion chamber is about 300 μs, and the time for the flame to reach a steady state is tens of milliseconds. When the ignition energy is increased by 1.5 times, the time required for the flame kernel to develop from the recirculation zone to the mainstream zone is shortened to 1/3 of the original; with the increase of the fuel-oxygen momentum ratio, the shape of the initial flame kernel changes from round to oval, and the development speed of the flame kernel also increases. The fuel-oxygen momentum ratio increases from 0.59 to 2.75, and the time required for the flame kernel to develop from the recirculation zone to the mainstream zone decreases from 167 μs to 56 μs＀ Except chamber pressure, other factors have no effect on the pressure rise rate of the ignition process, and the higher the chamber pressure, the faster the pressure rise rate; in order to avoid deflagration, the shear layer is more suitable as the ignition position. With the increase of ignition energy, fuel-oxygen momentum ratio and chamber pressure, the development speed of flame kernel increases, but the shape and development process of flame kernel will be different under the influence of different factors.

On the simulation of a wing with detached end-plates using an actuator line based on an anisotropic Gaussian kernel

The results of an actuator line method (ALM) using an anisotropic Gaussian kernel for both, velocity sampling and force projection, are compared to the wall-resolved results of a NACA 0015 rectangular wing at an angle of attack of 10°. The rectangular wing is simulated both, with and without, detached end-plates to show that the ALM is capable of accounting for the presence of a narrow gap of 0.01 c between the tip of the wing and the end-plate. The anisotropic kernel shape is varied in the chordwise, thickness-wise, and spanwise directions for different grid sizes and number of actuating points. The body forces are truncated and regularized to enforce proper spacing between the actuating line and the end-plate. ALM and wall-resolved results are compared on the basis of the integrated lift and drag coefficients, the sectional lift and drag coefficients, as well as the tip vortex core size, position, and circulation. Recommendations are made regarding the optimal kernel shape for a given number of actuating points and mesh resolution.

SOKS: Automatic Searching of the Optimal Kernel Shapes for Stripe-Wise Network Pruning.

In spite of the remarkable performance, deep convolutional neural networks (CNNs) are typically over-parameterized and computationally expensive. Network pruning has become a popular approach to reducing the storage and calculations of CNN models, which commonly prunes filters in a structured way or discards single weights without structural constraints. However, the redundancy in convolution kernels and the influence of kernel shapes on the performance of CNN models have attracted little attention. In this article, we develop a framework, termed searching of the optimal kernel shape (SOKS), to automatically search for the optimal kernel shapes and perform stripe-wise pruning (SWP). To be specific, we introduce coefficient matrices regularized by a variety of regularization terms to locate important kernel positions. The optimal kernel shapes not only provide appropriate receptive fields for each convolution layer, but also remove redundant parameters in convolution kernels. SWP is also achieved by utilizing these irregular kernels and actual inference speedups on the graphics processing unit (GPU) are obtained. Comprehensive experimental results demonstrate that SOKS searches high-efficiency kernel shapes and achieves superior performance in terms of both compression ratio and inference latency. Embedding the searched kernels into VGG-16 increases the accuracy from 93.53% to 94.26% on CIFAR-10, while pruning 59.27% model parameters and reducing 27.07% inference latency.

Rotation-equivariant correspondence matching based on a dual-activation mixer

Learning-based correspondence matching methods have become the mainstream techniques in many computer vision and robotics applications due to their robustness to large illumination and viewpoint changes. However, it is difficult for conventional convolutional neural networks (CNNs) to extract rotation-equivariant local features. Recent work has shown that CNNs combined with group-equivariant architectures are surprisingly effective at matching correspondences even when the images are rotated to a dramatic extent. However, the inherent shape (square) of convolution kernels causes the performance bottleneck of such rotation-equivariant CNNs. To address this issue, we propose an adaptive dual rotation-equivariant correspondence matching algorithm, which performs stably at all angles. We mathematically analyze the effectiveness of our proposed rotation-equivariant correspondence matching approach and its performance with respect to different convolution kernels. Extensive experiments on the Rotated-HPatches, SIM2E, and MegaDepth datasets demonstrate the effectiveness of our proposed algorithm.

Deep learning in computed tomography super resolution using multi-modality data training.

One of the limitations in leveraging the potential of artificial intelligence in X-ray imaging is the limited availability of annotated training data. As X-ray and CT shares similar imaging physics, one could achieve cross-domain data sharing, so to generate labeled synthetic X-ray images from annotated CT volumes as digitally reconstructed radiographs (DRRs). To account for the lower resolution of CT and the CT-generated DRRs as compared to the real X-ray images, we propose the use of super-resolution (SR) techniques to enhance the CT resolution before DRRgeneration. As spatial resolution can be defined by the modulation transfer function kernel in CT physics, we propose to train a SR network using paired low-resolution (LR) and high-resolution (HR) images by varying the kernel's shape and cutoff frequency. This is different to previous deep learning-based SR techniques on RGB and medical images which focused on refining the sampling grid. Instead of generating LR images by bicubic interpolation, we aim to create realistic multi-detector CT (MDCT) like LR images from HR cone-beam CT (CBCT)scans. We propose and evaluate the use of a SR U-Net for the mapping between LR and HR CBCT image slices. We reconstructed paired LR and HR training volumes from the same CT scans with small in-plane sampling grid size of . We used the residual U-Net architecture to train two models. SRUN : trained with kernel-based LR images, and SRUN : trained with bicubic downsampled data as baseline. Both models are trained on one CBCT dataset (n = 13 391). The performance of both models was then evaluated on unseen kernel-based and interpolation-based LR CBCT images (n = 10 950), and also on MDCT images (n = 1392). Five-fold cross validation and ablation study were performed to find the optimal hyperparameters. Both SRUN and SRUN models show significant improvements (p-value 0.05) in mean absolute error (MAE), peak signal-to-noise ratio (PSNR) and structural similarity index measures (SSIMs) on unseen CBCT images. Also, the improvement percentages in MAE, PSNR, and SSIM by SRUN is larger than SRUN . For SRUN , MAE is reduced by 14%, and PSNR and SSIMs increased by 6 and 8%, respectively. To conclude, SRUN outperforms SRUN , which the former generates sharper images when tested with kernel-based LR CBCT images as well as cross-modality LR MDCTdata. Our proposed method showed better performance than the baseline interpolation approach on unseen LR CBCT. We showed that the frequency behavior of the used data is important for learning the SR features. Additionally, we showed cross-modality resolution improvements to LR MDCT images. Our approach is, therefore, a first and essential step in enabling realistic high spatial resolution CT-generated DRRs for deep learningtraining.

Kernel adaptive memory network for blind video super-resolution

Although recent video super-resolution (VSR) works show remarkable restoration performance for low-resolution (LR) video downscaled by a fixed known blur kernel, blind VSR suffers from severe performance degradation when the blur kernel is unknown. To alleviate this problem, blur kernel estimation methods have been proposed for VSR. However, existing VSR models must be trained separately with each LR dataset downscaled using all possible blur kernels. This is a time-consuming and memory-consuming task. To address these issues, we propose a kernel adaptive memory network for a blind VSR (KeMoVSR). The KeMoVSR mainly consists of a dual regression blur kernel estimator and a kernel adaptive VSR. The proposed blur kernel estimator predicts the parametric and non-parametric blur kernels by exploiting the blur kernel variation. Due to the blur kernel variation, the proposed kernel estimator can consider the temporal consistency of the blur kernel variation in adjacent frames, which leads to accurate blur kernel estimation in LR video frames. The proposed kernel adaptive VSR modulates the VSR weights according to the shape of the blur kernel by weight modulation layers. By integrating the proposed methods based on the memory network, we propose the KeMoVSR, which performs VSR by adaptively modulating the VSR weights using the blur kernel parameters as keys and values in memory networks. Experiments show that the KeMoVSR achieves superior performance compared to other blind VSR approaches. The KeMoVSR provides effective memory utilization that is appropriate for real-world scenarios. The code is available at https://github.com/dbseorms16/KeMoVSR.

QGLF5 from Oryza rufipogon Griff. improves kernel shape, plant architecture, and yield in rice.

A novel QTL qGLF5 from Oryza rufipogon Griff. improves yield per plant and plant architecture in rice. Kernel size and plant architecture are critical agronomic traits that are key targets for improving crop yield. From the single-segment substitution lines of Oryza rufipogon Griff. in the indica cultivar Huajingxian74 (HJX74) background, we identified a novel quantitative trait locus (QTL), named qGLF5, which improves kernel shape, plant architecture, and yield per plant in rice. Compared with the control HJX74, the plant height, panicles per plant, panicle length, primary branches per panicle, secondary branches per panicle, and kernels per plant of the near-isogenic line-qGLF5 (NIL-qGLF5) are significantly increased. NIL-qGLF5 has long and narrow kernels by regulating cell number, cell length and width in the spikelet hulls. Yield per plant of NIL-qGLF5 is increased by 35.02% compared with that of HJX74. In addition, qGLF5 significantly improves yield per plant and plant architecture of NIL-gw5 and NIL-GW7. These results indicate that qGLF5 might be beneficial for improving plant architecture and kernel yield in rice breeding by molecular design.