Recently, referring image segmentation has aroused widespread interest. Previous methods perform the multi-modal fusion between language and vision at decoding side of network. And, linguistic feature interacts with visual each scale separately, which ignores continuous guidance to multi-scale features. In this work, we propose an encoder network (EFN), transforms into a learning network, uses refine features progressively. Moreover, co-attention mechanism is embedded in EFN realize parallel...

Most existing methods do not explicitly formulate the mutual guidance between vision and language. In this work, we propose a bi-directional relationship inferring network (BRINet) to model dependencies of cross-modal information. detail, vision-guided linguistic attention is used learn adaptive context corresponding each visual region. Combining with language-guided attention, module (BCAM) built multi-modal features. Thus, ultimate semantic target object referring expression can be...

Multi-hop reasoning over real-life knowledge graphs (KGs) is a highly challenging problem as traditional subgraph matching methods are not capable to deal with noise and missing information. Recently, address this promising approach based on jointly embedding logical queries KGs into low-dimensional space identify answer entities has emerged. However, existing proposals ignore critical semantic inherently available in KGs, such type To leverage information, we propose novel type-aware model,...

Hyper-relational knowledge graphs (HKGs) extend standard by associating attribute-value qualifiers to triples, which effectively represent additional fine-grained information about its associated triple. graph completion (HKGC) aims at inferring unknown triples while considering qualifiers. Most existing approaches HKGC exploit a global-level structure encode hyper-relational into the convolution message passing process. However, addition of multi-hop might bring noise triple prediction To...

Unmanned Aerial Vehicles (UAVs) often fly in formation owing to their lightness, flexibility and versatility, meaning that the distances between individual pairs of UAVs stay fixed. A four-level hybrid architecture is presented this paper, with mission planning level, manage control level UAV level. architecture, we can complete a typical scenario multi-UAV, which consists reaching, keeping, collision avoidance reconfiguration.

To explore the application of a traditional machine learning model in intelligent management pigs, this paper, influence PCA pre-treatment on pig face identification with RF is studied. By testing method, parameters two schemes, one adopting alone and other + PCA, were determined to be 65 70, respectively. With individual tests carried out 10 accuracy, recall, f1-score increased by 2.66, 2.76, 2.81 percentage points, Except for slight increase training time, test time was reduced 75% old...

<title>Abstract</title> The development of a bio-sensing strategy based on CRISPR/Cas that is exceptionally sensitive crucial for the identification trace molecules. Colorimetric miRNA detection utilizing CRISPR/Cas13a-triggered DNAzyme signal amplification was described in this article. developed implemented miRNA-21 as proof-of-concept. cleavage activity Cas13a triggered when target molecule bonded to Cas13a-crRNA complex and cleaved uracil ribonucleotides (rU) substrate probe. As...

Recently, referring image localization and segmentation has aroused widespread interest. However, the existing methods lack a clear description of interdependence between language vision. To this end, we present bidirectional relationship inferring network (BRINet) to effectively address challenging tasks. Specifically, first employ vision-guided linguistic attention module perceive keywords corresponding each region. Then, language-guided visual adopts learned adaptive guide update...

Referring image segmentation (RIS) has obtained an impressive achievement by fully convolutional networks (FCNs). However, previous RIS methods require a large number of pixel-level annotations. In this article, we present weakly supervised method using bounding box (BB) the first stage, introduce adversarial boundary loss to extract object contour from BB, which is then used select appropriate region proposals for pseudoground-truth (PGT) generation. second design co-training (Co-T)...

The metamodeling approach has been an important method to reduce the computational expense of complex system simulation. Metamodeling is process building a ldquomodel modelrdquo provide fast surrogate model for expensive simulation code. Main techniques include polynomial regression, kriging, radial basis function and support vector regression. In this paper we investigate relevance regression (RVR) as alternative To further understand new method, compare its performance with other four...

Recently, driven by numerous publicly available machine reading comprehension (MRC) datasets, MRC systems have made some progress.These however, two major limitations: 1) the defined tasks are relatively simple, and 2) they do not provide explainable evaluation which is critical to objectively comprehensively review reasoning capabilities of current systems.In this paper, we propose GCRC, a new dataset with challenging high-quality multi-choice questions, collected from Gaokao Chinese...

In view of the obvious differences in manifestations same diseases apples at different stages disease, show certain similarities, and early symptoms disease are not obvious. For these problems, a new model attention residual network (ARNet) was introduced based on combination thought. The introduces multi-layer modules to solve problems location dispersion features that difficult extract. order avoid degradation, module constructed effectively integrate high low-level features, data augment...

The area of the pig's face contains rich biological information, such as eyes, nose, and ear. high-precision detection pig postures is crucial to identification pigs, it can also provide fundamental archival information for study abnormal behavioral characteristics regularities. In this study, a series attention blocks were embedded in Feature Pyramid Network (FPN) automatic posture group-breeding environments. Firstly, Channel Attention Block (CAB) Position (PAB) proposed capture channel...

Individual identification and behavioural analysis of pigs is a key link in the intelligent management piggery, for which computer vision technology based on application improvement deep learning model has become mainstream. However, operation high requirements to hardwares, also weak interpretability, make it difficult adapt both mobile terminals embedded applications. In this study, first put forward that facial features can be extracted by PCA method before eigen face adopted verification...

Multimodal entity linking (MEL) aims to link ambiguous mentions within multimodal contexts corresponding entities in a knowledge base. Most existing approaches MEL are based on representation learning or vision-and-language pre-training mechanisms for exploring the complementary effect among multiple modalities. However, these methods suffer from two limitations. On one hand, they overlook possibility of considering negative samples same modality. other lack capture bidirectional cross-modal...