This paper focuses on the visible-thermal cross-modality person re-identification (VT Re-ID) task, whose goal is to match images between daytime visible modality and nighttime thermal modality. The two-stream network usually adopted address discrepancy, most challenging problem for VT Re-ID, by learning multi-modality features. In this paper, we explore how many parameters a should share, which still not well investigated in existing literature. By splitting ResNet50 model construct...

This letter presents a conceptually simple and effective dual-granularity triplet loss for visible-thermal person re-identification (VT-ReID). Generally, ReID models are always trained with the sample-based identification from fine granularity level. Further, center-based could be introduced to encourage intra-class compactness inter-class discrimination coarse Our proposed well organizes in hierarchical manner, just some configurations of typical operations, such as pooling batch...

Laryngeal cancer tumor (LCT) grading is a challenging task in P63 Immunohistochemical (IHC) histopathology images due to small differences between LCT levels pathology images, the lack of precision lesion regions interest (LROIs) and paucity image samples. The key solving problem transfer knowledge from other identify more accurate LROIs, but following problems occur: 1) transferring without priori experience often causes negative creates heavy workload abundance types, 2) convolutional...

This paper proposes a new method to identify local damages in frame structures based on approximate Metropolis–Hastings (AMH) algorithm and statistical moment. By analyzing the sensitivity of different moment-based damage indices, fusion index fourth-order displacement moment eighth-order acceleration is selected. Then are primarily evaluated by AMH algorithm, where Gibbs sampling adopted. Finally, uncertainty identified analyzed using probability density evolution (PDEM). Numerical...

Images taken on snowy days often suffer from severe negative visual effects caused by snowflakes. The task of removing snowflakes a image is known as desnowing, which challenging details are easily mistakenly treated and thus may be significantly lost during snowflake removal. Leveraging invertible neural networks (INNs), this paper presents deep learning-based method for single can remove accurately while preserving well. Interpreting desnowing an decomposition problem, we propose INN...

Cervical cancer is the fourth most common in world. Whole-slide images (WSIs) are an important standard for diagnosis of cervical cancer. Missed diagnoses and misdiagnoses often occur due to high similarity pathological images, large number readings, long reading time, insufficient experience levels pathologists. Existing models have feature extraction representation capabilities, they suffer from classification. Therefore, this work first designs image processing algorithm data...

This paper focuses on the high-resolution (HR) remote sensing images semantic segmentation task, whose goal is to predict labels in a pixel-wise manner. Due rich complexity and heterogeneity of information HR images, ability extract spatial details (boundary information) context dominates performance segmentation. In this paper, based frequently used fully convolutional network framework, we propose boundary enhancing (BES-Net) explicitly use enhance extraction. BES-Net mainly consists three...

Single-layer dual-/tri-band substrate integrated waveguide (SIW) filtenna based on multifunctional cavity-backed slots is proposed in this communication. The etched the SIW cavity can be considered as radiators well resonators. When a pair of back-to-back open-loop embedded with straight slot cavity, three resonant modes generated by combine TE101 mode to form superior dual-band filtering response. Moreover, four radiation nulls (RNs) located at both sides two operational passbands are...

Salient object detection (SOD) aims to identify the most prominent regions in images. However, large model sizes, high computational costs, and slow inference speeds of existing RGB-D SOD models have hindered their deployment on real-world embedded devices. To address this issue, we propose a novel method named AirSOD, which is committed lightweight SOD. Specifically, first design hybrid feature extraction network, includes three stages MobileNetV2 our Parallel Attention-Shift convolution...

The performance of a deep learning-based Synthetic aperture radar (SAR) automatic target recognition (ATR) model largely relies on the scale and quality training samples. However, it is time-consuming expensive to collect sufficient data in practice. Although generative adversarial network (GAN) provides way for SAR image generation, existing GAN-based methods cannot confirm what features generator learns, thus they struggle generating precise images. In this paper, we propose an angle...

Electronic tongue (E-Tongue), as a novel taste analysis tool, shows promising perspective for recognition. In this paper, we constructed voltammetric E-Tongue system and measured 13 different kinds of liquid samples, such tea, wine, beverage, functional materials, etc. Owing to the noise variety environmental conditions, acquired data inseparable patterns. To end, from viewpoint algorithm, propose local discriminant preservation projection (LDPP) model, an under-studied subspace learning...

The results of the linear range cell migration (RCM) correction and inherent range-dependent squint angle in case high-resolution highly squinted synthetic aperture radar (SAR) imaging produce two-dimensional (2-D) spatial-variant RCMs azimuth-dependent Doppler parameters (i.e., varying centroid frequency modulation rates), which make SAR difficult. However, most existing algorithms failed to consider these problems. To obtain high-quality image, this study, both 2-D are studied. First, a...

An innovative iterative process is proposed to acquire the dynamic of multichannel slotted ALOHA (S-ALOHA). It reveals direct relation between number contending devices that perform their jth random access (RA) attempt at ith RA slot and newly arrived before slot. These results allow engineers analytically derive probability density function delay S-ALOHA, as well its cumulative average value. Under stable attempts assumption, simplified form above analysis given, with which we prove...

Representation-based target detectors for hyperspectral imagery (HSI) have recently aroused a lot of interests. However, existing methods ignore the dictionary structure and cannot guarantee an informative discriminative representation test pixels detection. To alleviate problem, this letter proposes novel sparse dense hybrid representation-based detector (SDRD). The proposed adopts idea that relationship between background sub-dictionaries is collaborative competition. discovered preserved...

Change detection is one of the fundamental applications synthetic aperture radar (SAR) images. However, speckle noise presented in SAR images has a negative effect on change detection, leading to frequent false alarms mapping products. In this research, novel two-phase object-based deep learning approach proposed for multi-temporal image detection. Compared with traditional methods, brings two main innovations. One classify all pixels into three categories rather than categories: unchanged...

In recent years, with the rise of intelligent hardware technology, Internet Things (IoT) has achieved rapid development and been widely used in smart cities, power grids, industrial Internet, other fields. The resulting security risks IoT have attracted more attention. Being different from traditional authentication that is based on MAC or certificate, RF fingerprinting technology extracts fingerprints emissions wireless transmitters. These root imperfection transmitting circuits can be for...

Existing sparsity-based hyperspectral image (HSI) target detection methods have two key problems. 1) The background dictionary is locally constructed by the pixels between inner and outer windows, surrounding enclosing central test pixel. dual-window strategy intricate might result in impure deteriorating performance. 2) For an unbalanced binary classification problem, atoms are generally inadequate compared with dictionary, which yield unstable issues, this article proposes a novel...

From image to video understanding, the capabilities of Multi-modal LLMs (MLLMs) are increasingly powerful. However, most existing understanding benchmarks relatively short, which makes them inadequate for effectively evaluating long-sequence modeling MLLMs. This highlights urgent need a comprehensive and integrated long benchmark assess ability MLLMs thoroughly. To this end, we propose ALLVB (ALL-in-One Long Video Understanding Benchmark). ALLVB's main contributions include: 1) It integrates...