Existing approaches for replay and synthetic speech detection still lack generalizability to unseen spoofing attacks. This work proposes leverage a novel model structure, so-called Res2Net, improve the anti-spoofing countermeasure’s generalizability. Res2Net mainly modifies ResNet block enable multiple feature scales. Specifically, it splits maps within one into channel groups designs residual-like connection across different groups. Such increases possible receptive fields, resulting in...

To address the threat of drones intruding into high-security areas, real-time detection is urgently required to protect these areas. There are two main difficulties in drones. One them that move quickly, which leads requiring faster detectors. Another problem small difficult detect. In this paper, firstly, we achieve high accuracy by evaluating three state-of-the-art object methods: RetinaNet, FCOS, YOLOv3 and YOLOv4. Then, first problem, prune convolutional channel shortcut layer YOLOv4...

Sink mobility has attracted much research interest in recent years because it can improve network performance such as energy efficiency and throughput. An energy-unconscious moving strategy is potentially harmful to the balance of consumption among sensor nodes so aggravate hotspot problem networks. In this paper, we propose an autonomous for mobile sinks data-gathering applications. our solution, a sink approaches with high residual force them forward data other tries avoid passing by low...

Few-shot semantic segmentation (FSS) aims to segment novel classes with only a few annotated samples. Existing methods FSS generally combine the mask and corresponding support image generate class-specific representation, perform for query by matching features of these representations. However, performance could be fragile lack an effective method handle inappropriate use neglection correlation between in images. In this work, we propose Disentanglement Recombination Network (DRNet)...

High Efficiency Video Coding (HEVC) INTRA coding improves compression efficiency by adopting advanced technologies, such as multi-level quad-tree block partitioning and up to 35-mode prediction. However, it significantly increases the complexity, memory access, power consumption, which goes against its widely applications, especially for ultra-high definition and/or mobile video applications. To tackle this problem, we propose effective data driven unit (CU) size decision approaches HEVC...

Underwater image datasets are crucial in underwater vision research. Because of the strong absorption and scattering effects that occur underwater, some ground truth such as depth map, which can be easily collected in-air, becomes a great challenge environments. To solve issues associated with lack truth, we propose trainable end-to-end system an multistyle generative adversarial network (UMGAN) takes advantage cycle-consistent (CycleGAN) conditional networks. This generate multiple...

Chromophoric dissolved organic matter (CDOM) is crucial in the biogeochemical cycle and carbon of aquatic environments. However, inland waters, remotely sensed estimates CDOM remain challenging due to low optical signal complex conditions. Therefore, developing efficient, practical robust models estimate absorption coefficient waters essential for successful water environment monitoring management. We examined improved different machine learning algorithms using extensive measurements...

In this article, we propose a quality assessment model-based on the projection invariant feature and visual saliency for Stereoscopic Omnidirectional Images (SOIs). Firstly, monocular binocular features of SOI are derived from Scale-Invariant Feature Transform (SIFT) points to tackle inconsistency between stretched formats viewports. Secondly, model, which combines chrominance contrast perceptual factors, is used facilitate prediction accuracy. Thirdly, according characteristics panoramic...

The median-type filter is an effective technique to remove salt and pepper (SAP) noise; however, such a mechanism cannot always effectively noise preserve details due the local diversity singularity non-stationarity. In this paper, two-step SAP removal method was proposed based on analysis of errors. first step, used process image corrupted by noise. Then, in second novel-designed adaptive nonlocal bilateral weaken error filter. By building histograms errors, we found that almost obeys...

Driving fatigue is the cause of many traffic accidents and poses a serious threat to road safety. To address this issue, paper aims develop system for early detection driver fatigue. The leverages heart rate variability (HRV) features embedded machine learning estimate driver’s level. HRV derived from electrocardiogram (ECG) signals captured by wearable device analysis. Time- frequency-domain are then extracted used as input classifier. A dataset collected driving simulation experiment...

ABSTRACT Salt‐alkali stress is one of the most widespread and devastating abiotic stress. Alternative splicing a response pathway to such However, role microexons in salt‐alkali soybean remains obscure. In this study, we identified related We focused on analyzing conserved sequence patterns 27–30 bp microexons, consistently observed GT AG sequences at 5′ 3′ ends these microexons. Additionally, found that AP2 protein domain had abundant Interestingly, majority transcription factor were 9...

For the clearing of hazy images, it is difficult to obtain dehazing datasets with paired mapping images. Currently, most algorithms are trained on synthetic insufficient complexity, which leads model overfitting. At same time, physical characteristics fog in real world ignored current algorithms; that is, degree related depth field and scattering coefficient. Moreover, only consider image land scenes ignore maritime scenes. To address these problems, we propose a multi-scene algorithm based...

Deep learning models have demonstrated their potential in effective molecular representations critical for drug property prediction and discovery. Despite significant advancements leveraging multimodal molecule semantics, existing approaches often struggle with challenges such as low-quality data structural complexity. Large language (LLMs) excel generating high-quality due to robust characterization capabilities. In this work, we introduce GICL, a cross-modal contrastive framework that...

In this paper, the formulation of Stolt migration is modified for impulse borehole radar near-field imaging in subsurface scenarios where transceiver widely separated with respect to detection range. The proposed approach consists following aspects. First, locations transmitter and receiver survey are regarded as independent sample dimensions, original set converted an enlarged virtual set. frequency-wavenumber spectrum (FWS) available via multidimensional fast Fourier transform (FFT). Then,...

In this article, statistical Early Termination (ET) and Skip (ES) models are proposed for fast Coding Unit (CU) prediction mode decision in HEVC INTRA coding, which three categories of ET ES sub-algorithms included. First, the CU ranges current recursively predicted based on texture depth spatial neighboring CUs. Second, model schemes applied to optimize decision, coding complexities over different layers jointly minimized subject acceptable rate-distortion degradation. Third, correlations...

Azimuth gamma logging while drilling (LWD) is one of the important technologies geosteering but information real-time data transmission limited and interpretation difficult. This study proposes a method applying artificial intelligence in LWD to enhance accuracy efficiency processing. By examining formation response characteristics azimuth ray (GR) curve, preliminary change position detected based on wavelet transform modulus maxima (WTMM) method, then dynamic threshold determined, set...

Dynamic point cloud is a volumetric visual data representing realistic 3D scenes for virtual reality and augmented applications. However, its large volume has been the bottleneck of processing, transmission, storage, which requires effective compression. In this paper, we propose Perceptually Weighted Rate-Distortion Optimization (PWRDO) scheme Video-based Point Cloud Compression (V-PCC), aims to minimize perceptual distortion reconstructed at given bit rate. Firstly, general framework...

This study focuses on optimizing precipitation forecast induced by tropical cyclones (TCs) in the Northwest Pacific region, with lead times ranging from 6 to 72 h. The research employs deep learning models, such as U-Net, UNet3+, SE-Net, and SE-UNet3+, which utilize data Global Forecast System (GFS) real-time GFS environmental background using a U-Net structure. To comprehensively make use of forecasts these we additionally probabilistic matching (PM) simple averaging (AVR) rainfall...