In this paper, we propose a new clustering model, called DEeP Embedded Regularized ClusTering (DEPICT), which efficiently maps data into discriminative embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of multinomial logistic regression function stacked on top multi-layer convolutional autoencoder. We define objective using relative entropy (KL divergence) minimization, regularized by prior for the frequency An alternating strategy is then derived to...

Thanks to the success of deep learning, cross-modal retrieval has made significant progress recently. However, there still remains a crucial bottleneck: how bridge modality gap further enhance accuracy. In this paper, we propose self-supervised adversarial hashing (SSAH) approach, which lies among early attempts incorporate learning into in fashion. The primary contribution work is that two networks are leveraged maximize semantic correlation and consistency representations between different...

Most existing Re-IDentification (Re-ID) methods are highly dependent on precise bounding boxes that enable images to be aligned with each other. However, due the challenging practical scenarios, current detection models often produce inaccurate boxes, which inevitably degenerate performance of Re-ID algorithms. In this paper, we propose a novel coarse-to-fine pyramid model relax need not only incorporates local and global information, but also integrates gradual cues between them. The is...

Given the benefits of its low storage requirements and high retrieval efficiency, hashing has recently received increasing attention. In particular, cross-modal been widely successfully used in multimedia similarity search applications. However, almost all existing methods employing cannot obtain powerful hash codes due to their ignoring relative between heterogeneous data that contains richer semantic information, leading unsatisfactory performance. this paper, we propose a tripletbased...

With benefits of low storage cost and fast query speed, cross-modal hashing has received considerable attention recently. However, almost all existing methods on cannot obtain powerful hash codes due to directly utilizing hand-crafted features or ignoring heterogeneous correlations across different modalities, which will greatly degrade the retrieval performance. In this paper, we propose a novel deep method generate compact through an end-to-end learning architecture, can effectively...

The clustering methods have recently absorbed even-increasing attention in learning and vision. Deep combines embedding together to obtain optimal subspace for clustering, which can be more effective compared with conventional methods. In this paper, we propose a joint framework discriminative spectral clustering. We first devise dual autoencoder network, enforces the reconstruction constraint latent representations their noisy versions, embed inputs into space As such learned robust noise....

Referring expression comprehension (REC) and segmentation (RES) are two highly-related tasks, which both aim at identifying the referent according to a natural language expression. In this paper, we propose novel Multi-task Collaborative Network (MCN) achieve joint learning of REC RES for first time. MCN, can help better language-vision alignment, while locate referent. addition, address key challenge in multi-task setup, i.e., prediction conflict, with innovative designs namely, Consistency...

The success of convolutional neural networks (CNNs) in computer vision applications has been accompanied by a significant increase computation and memory costs, which prohibits their usage on resource-limited environments, such as mobile systems or embedded devices. To this end, the research CNN compression recently become emerging. In paper, we propose novel filter pruning scheme, termed structured sparsity regularization (SSR), to simultaneously speed up reduce overhead CNNs, can be well...

Hashing plays a pivotal role in nearest-neighbor searching for large-scale image retrieval. Recently, deep learning-based hashing methods have achieved promising performance. However, most of these involve discriminative models, which require large-scale, labeled training datasets, thus hindering their real-world applications. In this paper, we propose novel strategy to exploit the semantic similarity data and design an efficient generative adversarial framework learn binary hash codes...

Owing to the advantages of low storage cost and high query efficiency, cross-modal hashing has received increasing attention recently. As failing bridge inherent modality gap between modalities, most existing methods have limited capability explore semantic consistency information different data, leading unsatisfactory search performance. To address this problem, we propose a novel deep method named Multi-Task Consistency-Preserving Adversarial Hashing (CPAH) fully correlation modalities for...

Due to storage and search efficiency, hashing has become significantly prevalent for nearest neighbor search. Particularly, deep methods have greatly improved the performance, typically under supervised scenarios. In contrast, unsupervised models can hardly achieve satisfactory performance due lack of supervisory similarity signals. To address this problem, in paper, we propose a new model, called DistilHash, which learn distilled data set, where pairs confident Specifically, investigate...

By contrasting positive-negative counterparts, graph contrastive learning has become a prominent technique for unsupervised representation learning. However, existing methods fail to consider the class information and will introduce false-negative samples in random negative sampling, causing poor performance. To this end, we propose debiased framework, which can jointly perform clustering. Specifically, representations be optimized by aligning with clustered information, simultaneously,...

The privacy-preserving federated learning for vertically partitioned (VP) data has shown promising results as the solution of emerging multiparty joint modeling application, in which holders (such government branches, private finance, and e-business companies) collaborate throughout process rather than relying on a trusted third party to hold data. However, most existing algorithms VP are limited synchronous computation. To improve efficiency when unbalanced computation/communication...

Graph matching, which refers to a class of computational problems finding an optimal correspondence between the vertices graphs minimize (maximize) their node and edge disagreements (affinities), is fundamental problem in computer science relates many areas such as combinatorics, pattern recognition, multimedia vision. Compared with exact graph (sub)isomorphism often considered theoretical setting, inexact weighted matching receives more attentions due its flexibility practical utility. A...

Hashing is becoming increasingly popular for approximate nearest neighbor searching in massive databases due to its storage and search efficiency. Recent supervised hashing methods, which usually construct semantic similarity matrices guide hash code learning using label information, have shown promising results. However, it relatively difficult capture utilize the relationships between points unsupervised settings. To address this problem, we propose a novel deep framework called Semantic...

With explosive growth of data volume and ever-increasing diversity modalities, cross-modal similarity search, which conducts nearest neighbor search across different has been attracting increasing interest. This paper presents a deep compact code learning solution for efficient search. Many recent studies have proven that quantization-based approaches perform generally better than hashing-based on single-modal In this paper, we propose quantization approach, is among the early attempts...

Robust reversible watermarking (RRW) methods are popular in multimedia for protecting copyright, while preserving intactness of host images and providing robustness against unintentional attacks. However, conventional RRW not readily applicable practice. That is mainly because 1) they fail to offer satisfactory reversibility on large-scale image datasets; 2) have limited extracting watermarks from the watermarked destroyed by different attacks; 3) some them suffer extremely poor invisibility...

Feature-based image watermarking schemes, which aim to survive various geometric distortions, have attracted great attention in recent years. Existing schemes shown robustness against rotation, scaling, and translation, but few are resistant cropping, nonisotropic random bending attacks (RBAs), affine transformations. Seo Yoo present a geometrically invariant based on covariant regions (ACRs) that provide certain degree of robustness. To further enhance the robustness, we propose new scheme...

Hashing technique has become a promising approach for fast similarity search. Most of existing hashing research pursue the binary codes same type entities by preserving their similarities. In practice, there are many scenarios involving nearest neighbor search on data given in matrix form, where two different types of, yet naturally associated respectively correspond to its dimensions or views. To fully explore duality between views, we propose collaborative scheme form enable various...

Face alignment has been extensively studied in computer vision community due to its fundamental role facial analysis, but it remains an unsolved problem. The major challenges lie the highly nonlinear relationship between face images and associated shapes, which is coupled by underlying correlation of landmarks. Existing methods mainly rely on cascaded regression, suffering from intrinsic shortcomings, e.g., strong dependency initialization failure exploit landmark correlations. In this...

SUMMARY Zinc (Zn) is an essential micronutrient for most organisms including humans, and Zn deficiency widespread in human populations, particularly underdeveloped regions. Cereals such as rice ( Oryza sativa ) are the major dietary source of people. However, molecular mechanism underlying uptake still not fully understood. Here, we report that a member ZIP (ZRT, IRT‐like protein) family, OsZIP9, contributes to rice. It was expressed epidermal exodermal cells lateral roots, localized plasma...

Arsenic (As) is toxic to organisms, and elevated As accumulation in rice (Oryza sativa) grain may pose a significant health risk humans. The predominant form of soil under aerobic conditions As(V), which has chemical structure similar that PO43-. Rice roots take up As(V) by phosphate (Pi) transporters, such as OsPT1 OsPT8. In the present study, we investigated contribution OsPT4, belonging Pht1 family, on uptake transport. We determined mRNA amounts OsPTs seedlings, expressions OsPT1, OsPT8...

Unsupervised deep hash functions have not shown satisfactory improvements against their shallow alternatives, and usually require supervised pretraining to avoid overfitting. In this paper, we propose a new unsupervised hashing function, called HashGAN, which efficiently obtains binary representation of input images without any pretraining. HashGAN consists three networks, generator, discriminator an encoder. By sharing the parameters encoder discriminator, benefit from adversarial loss as...