Hashing Methods Research Articles

Supervised Discrete Multiple-Length Hashing for Image Retrieval

Hashing can facilitate efficient retrieval and storage for large-scale images due to the binary representation. In the real applications, the trade-off between retrieval accuracy and speed is essential for designing a hashing framework, which is reflected by variable hash code lengths. In light of this, the existing hashing methods need to train different models for different lengths of hash codes, leading to considerable training time cost and hashing flexibility reduction. Given that a sample can be represented by various hash codes with different lengths, there are some helpful relationships that can boost the performance of hashing methods. However, the existing hashing methods do not fully utilize these relationships. To address the aforementioned issues, we propose a new model, known as supervised discrete multiple-length hashing (SDMLH), to simultaneously learn hash codes with multiple lengths. In this proposed SDMLH method, three types of information are respectively derived, from the hash codes with different lengths. The original features of the samples, and the label, are applied for hash learning. Unlike the existing hashing methods, SDMLH can fully employ the assistance among hash codes with different lengths and learn them in one step. Furthermore, given a hash length meeting the demand of users, we propose a hash fusion strategy to obtain the hash code with this desirable length by fusing the multiple-length hash codes. This obtained hash code outperforms the one learned directly. In addition, SDMLH can generate the hash code of any length that is shorter than the sum length of given multiple hash codes with the fusion strategy. To the best of our knowledge, SDMLH is one of the first attempts for learning multiple-length hash codes simultaneously. We conduct extensive experiments based on three benchmark datasets, demonstrating the superiority of this proposed method.

Refined Earthquake Focal Mechanism Catalog for Southern California Derived With Deep Learning Algorithms

AbstractEarthquake focal mechanisms, determined with P‐wave polarities and S/P amplitude ratios, are primary data for analyzing fault zone geometry, sense of slip, and the crustal stress field. Solving for the focal mechanisms of small earthquakes is often challenging because phase arrivals and first‐motion polarities are hard to be separated from noise. To overcome this challenge, we implement convolutional‐neural‐network algorithms (Ross, Meier, & Hauksson, 2018, Ross, Meier, Hauksson, & Heaton, 2018, https://doi.org/10.1029/2017jb015251, https://doi.org/10.1785/0120180080) to detect additional phases and polarities. Using both existing and these new data, we build a high‐quality focal mechanism catalog of 297,478 events that occurred from 1981 to 2021 in southern California with the HASH method of Hardebeck and Shearer (2002), https://doi.org/10.1785/0120010200, Hardebeck and Shearer (2003), https://doi.org/10.1785/0120020236. The new focal mechanism catalog is overall consistent with the standard catalog (Yang et al., 2012, https://doi.org/10.1785/0120110311) but includes 40% more focal mechanisms, and is more consistent with moment tensor solutions derived using waveform‐fitting methods. We apply the new catalog to identify changes in focal mechanism properties caused by the occurrences of large mainshocks such as the 2010 Mw7.2 El Mayor‐Cucapah and 2019 Mw7.1 Ridgecrest earthquakes. Such changes may be associated with co‐seismic stress drops, post‐seismic deformation processes, and static stress changes on a regional scale. The new high‐resolution catalog will contribute to improved understanding of the crustal stress field, earthquake triggering mechanisms, fault zone geometry, and sense of slip on the faults in southern California.

English

When new data streaming arrives, traditional hashing methods should retrain the hashing functions based on all samples. That leads to high training time complexity. In contrast, the online hashing algorithm re-computes the hashing functions just based on the new arrival streaming data and has been widely applied in large-scale image retrieval tasks. However, differences exist in numbers and labels between new arrival and old datasets, which causes the data imbalance problem while establishing their similarity matrix. This paper proposes a novel supervised online hashing method, Label Projection, based on Hadamard Codes for Online Hashing (LHOH), which jointly employs the label projection and similarity preservation mechanism to solve the data imbalance problem. In addition, LHOH considers the Hadamard codes as the label projection target domains to avoid the problem of difficult discrete optimization of the objective function. Then, LHOH employs the label projection matrix as label weight values, which can solve the data imbalance problem while computing the similarity matrix between new arrival and old datasets and preserve the consistency of Hamming and semantic space similarity. To increase the distinguishability among the hash codes, LHOH designs triple supervision learning mechanisms, including assigning Hadamard codes, projecting labels, and embedding labels. To validate the performance of the proposed LHOH method, this paper sets up the approximate nearest neighbor (ANN) search comparative experiments on two widely used datasets. The final results show that LHOH outperforms six current state-of-the-art online methods.

Interpretable pap-smear image retrieval for cervical cancer detection with rotation invariance mask generation deep hashing

Cervical cancer is a common disease in women, affecting their lives negatively and often resulting in death. Pap-smear tests are preferred by doctors as the primary tool in the early diagnosis and treatment of the disease. Physicians can be facilitated in the detection of five different categories of cervical cancer and similar cellular disease cases with the Pap-smear image retrieval technology. In this study, an algorithm for retrieval of cervical cancer images using hash coding with a Convolutional Neural Network (CNN) has been implemented. A sensitive deep hashing method combining interpretable mask generation and rotation invariance is proposed for cervical cancer detection. The distinctive features of cervical cancer cells with complex morphological features are focused on with the proposed hybrid dilated convolution spatial attention module and insignificant features are eliminated. Moreover, the loss function of Cauchy rotation invariance in terms of cervical cancer cell target is presented. In this way, the differences in the input samples are revealed, allowing the CNN to learn from different angles and achieve certain rotation invariance. The versatility and performance of the proposed method, as well as the efficiency of the loss function, have been tested on the SIPaKMeD and Mendeley LBC datasets consisting of cervical cancer images. In the experimental results obtained, it is shown that the proposed spatial attention module and rotational invariance deep hashing network generate high performance in cervical cancer image retrieval problems.

Predicting positions of flipped bits in robust image hashes

Both robust and cryptographic hash methods have advantages and disadvantages. It would be ideal if robustness and cryptographic confidentiality could be combined. The problem here is that the concept of similarity of robust hashes cannot be applied to cryptographic hashes. Therefore, methods must be developed to reliably intercept the degrees of freedom of robust hashes before they are included in a cryptographic hash, but without losing their robustness. To achieve this, we need to predict the bits of a hash that are most likely to be modified, for example after a JPEG compression. We show that machine learning can be used to make a much more reliable prediction than the approaches previously discussed in the literature.

BLEND: a fast, memory-efficientand accurate mechanism to find fuzzy seed matches in genome analysis.

Generating the hash values of short subsequences, called seeds, enables quickly identifying similarities between genomic sequences by matching seeds with a single lookup of their hash values. However, these hash values can be used only for finding exact-matching seeds as the conventional hashing methods assign distinct hash values for different seeds, including highly similar seeds. Finding only exact-matching seeds causes either (i)increasing the use of the costly sequence alignment or (ii)limited sensitivity. We introduce BLEND, the first efficient and accurate mechanism that can identify both exact-matching and highly similar seeds with a single lookup of their hash values, called fuzzy seed matches. BLEND (i) utilizes a technique called SimHash, that can generate the same hash value for similar sets, and (ii) provides the proper mechanisms for using seeds as sets with the SimHash technique to find fuzzy seed matches efficiently. We show the benefits of BLEND when used in read overlapping and read mapping. For read overlapping, BLEND is faster by 2.4×-83.9× (on average 19.3×), has a lower memory footprint by 0.9×-14.1× (on average 3.8×), and finds higher quality overlaps leading to accurate de novo assemblies than the state-of-the-art tool, minimap2. For read mapping, BLEND is faster by 0.8×-4.1× (on average 1.7×) than minimap2. Source code is available at https://github.com/CMU-SAFARI/BLEND.

Full-field deformation measurement of structural nodes based on panoramic camera and deep learning-based tracking method

Full-field deformation measurement of structures generally requires the aid of complex and expensive multi-camera measurement systems. A full-field structural deformation measurement method using a single panoramic camera and deep learning-based tracking algorithm is proposed. The contributions are as follows: (1) To address the problem that existing full-field image acquisition methods rely on multi-camera systems, a full-field image acquisition method based on a single panoramic camera is proposed, in which a distortion-free planar image covering the full-field of the structure is obtained by decomposing the projection in multiple directions based on the panoramic camera imaging model and the cubic projection method. (2) To solve the problem that the nodes of structures usually contain little texture and are difficult to track robustly with existing image processing methods, an object detection network with a modified tiny feature map layer and attention mechanism is applied to extract the region of interest (ROI) of each node automatically. (3) Finally, the ROIs of the identified nodes are clustered using a perceptual hashing method and then the node coordinates are calculated using a line segment detector (LSD) and intersection fitting. The proposed method is validated on a scaled model of a stadium, and the comparison to the deformation results of total station shows that the proposed method can calculate the displacement of all node at once, and the average error between the displacement results and those of the total station is 3.7 mm, which proves the practicality of the proposed method.

Deep Uncoupled Discrete Hashing via Similarity Matrix Decomposition

Hashing has been drawing increasing attention in the task of large-scale image retrieval owing to its storage and computation efficiency, especially the recent asymmetric deep hashing methods. These approaches treat the query and database in an asymmetric way and can take full advantage of the whole training data. Though it has achieved state-of-the-art performance, asymmetric deep hashing methods still suffer from the large quantization error and efficiency problem on large-scale datasets due to the tight coupling between the query and database. In this article, we propose a novel asymmetric hashing method, called D eep U ncoupled D iscrete H ashing (DUDH), for large-scale approximate nearest neighbor search. Instead of directly preserving the similarity between the query and database, DUDH first exploits a small similarity-transfer image set to transfer the underlying semantic structures from the database to the query and implicitly keep the desired similarity. As a result, the large similarity matrix is decomposed into two relatively small ones and the query is decoupled from the database. Then both database codes and similarity-transfer codes are directly learned during optimization. The quantization error of DUDH only exists in the process of preserving similarity between the query and similarity-transfer set. By uncoupling the query from the database, the training cost of optimizing the CNN model for the query is no longer related to the size of the database. Besides, to further accelerate the training process, we propose to optimize the similarity-transfer codes with a constant-approximation solution. In doing so, the training cost of optimizing similarity-transfer codes can be almost ignored. Extensive experiments on four widely used image retrieval benchmarks demonstrate that DUDH can achieve state-of-the-art retrieval performance with remarkable training cost reduction (30× - 50× relative).

METHOD OF LOW-RESOURCE HASHING TYPE "DATA – GENERATOR"

Creating a secure and efficient structure of a cryptographic algorithm is one of the key cryptographic tasks. Recently, cryptography for low-resource devices has attracted considerable attention of world scientists. A significant portion of the research is dedicated to the examination of block encryption methods, and conversely, there are notably fewer publicly disclosed proposals for low-resource hashing methods. Many hash functions recommended for use in low-resource devices by well-known standardization organizations are based on block encryption, which offers a sufficient level of security but demands substantial computational resources—something critical for their application in such devices. The urgency of investigating low-resource data hashing methods stems from the need to guarantee an adequate level of hash function security while minimizing computational resource usage through adjustments to the hashing process. This article reviews established approaches to constructing hash functions of varying complexities and examines the latest research and publications focused on low-resource hashing. Based on this, the structure and approach for developing a low-resource data hashing method were chosen. A novel low-resource hashing method, founded on the Merkle-Damgård construction and utilizing an iterative byte-oriented approach, is introduced. The process of low-resource hashing, according to the new method, is formally described. Statistical testing of the proposed method was conducted in accordance with NIST SP 800-22. An overview of the hardware implementation of the proposed low-resource hashing method is presented in the form of a generalized structural diagram. The complexity of the proposed hardware implementation is quantified in conventional units [GE] for hash value calculations of 128, 192, and 256 bits. A comparison of the proposed "data-generator" type hashing method with established low-resource hash functions, in terms of hardware costs, is conducted.

Non-Relaxation Deep Hashing Method for Fast Image Retrieval

Deep hashing methods utilize an end-to-end framework to mutually learn feature representations and hash codes, thereby achieving a better retrieval performance. Traditional supervised hashing methods adopt handcrafted features for hashing function learning and then generate hash codes through classification and quantization. The lack of adaptability and independence of the quantization procedure leads to low retrieval accuracy of supervised hashing methods with handcrafted features in image retrieval. In this study, a non-relaxation deep hashing method for fast image retrieval is proposed. In this method, a differentiable host thresholding function is used to encourage hash-like codes to approach -1 or 1 non-linearly at the output of the convolutional neural, instead of the symbol function for quantization used in the traditional method. The output of the host thresholding function is directly used to compute the network training error, and a loss function is elaborately designed with the norm to constrain each bit of the hash-like code to be as binary as possible. Finally, a symbol function is added outside the trained network model to generate binary hash codes for image storage and retrieval in a low-dimensional binary space. Extensive experiments on two large-scale public datasets show that our method can effectively learn image features, generate accurate binary hash codes, and outperform state-of-the-art methods in terms of the mean average precision.

ViT2CMH: Vision Transformer Cross-Modal Hashing for Fine-Grained Vision-Text Retrieval

In recent years, the development of deep learning has further improved hash retrieval technology. Most of the existing hashing methods currently use Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) to process image and text information, respectively. This makes images or texts subject to local constraints, and inherent label matching cannot capture fine-grained information, often leading to suboptimal results. Driven by the development of the transformer model, we propose a framework called ViT2CMH mainly based on the Vision Transformer to handle deep Cross-modal Hashing tasks rather than CNNs or RNNs. Specifically, we use a BERT network to extract text features and use the vision transformer as the image network of the model. Finally, the features are transformed into hash codes for efficient and fast retrieval. We conduct extensive experiments on Microsoft COCO (MS-COCO) and Flickr30K, comparing with baselines of some hashing methods and image-text matching methods, showing that our method has better performance.

Deep Hashing With Multi-Central Ranking Loss for Multi-Label Image Retrieval

With the explosive growth of various images, large-scale image retrieval has attracted ever-growing attention. Deep hashing methods have achieved great success on single-label retrieval. However, the multi-level similarities between samples in multi-label scenarios have not been fully explored. In this letter, based on the in-depth analysis of the complex semantic similarities of multi-label images, the Multi-Central Ranking Loss (MCR Loss) for deep hashing is proposed to construct a powerful metric space that not only preserves the fine-grained similarities of multi-label images but also has low quantization error. The proposed MCR Loss utilizes learnable hash centers and similarities of data-to-data pairs to optimize the metric space, which greatly alleviates the embedding conflict caused by proxy-based supervision, and reduces the quantization error. The proposed method is compared with several existing state-of-the-art hashing methods on two public multi-label benchmarks. Experimental results show that the proposed method achieves state-of-the-art performance on several ranking evaluation metrics.

Semi-Supervised Knowledge Distillation for Cross-Modal Hashing

Deep hashing methods have achieved tremendous success in cross-modal retrieval, due to its low storage consumption and fast retrieval speed. Supervised cross-modal hashing methods have achieved substantial advancement by incorporating semantic information. However, to a great extent, supervised methods rely on large-scale labeled cross-modal training data which are laborious to obtain. Moreover, most cross-modal hashing methods only handle two modalities of image and text, without taking the scene of multiple modalities into consideration. In this paper, we propose a novel semi-supervised approach called semi-supervised knowledge distillation for cross-modal hashing (SKDCH) to overcome the above-mentioned challenges, which enables guiding a supervised method using outputs produced by a semi-supervised method for multimodality retrieval. Specifically, we utilize teacher-student optimization to propagate knowledge. Furthermore, we improves triplet ranking loss to better mitigate the heterogeneity gap, which increases the discriminability of our proposed approach. Extensive experiments executed on two benchmark datasets validate that the proposed SKDCH surpasses the state-of-the-art methods.

Multi-Label Hashing for Dependency Relations among Multiple Objectives.

Learning hash functions have been widely applied for large-scale image retrieval. Existing methods usually use CNNs to process an entire image at once, which is efficient for single-label images but not for multi-label images. First, these methods cannot fully exploit independent features of different objects in one image, resulting in some small object features with important information being ignored. Second, the methods cannot capture different semantic information from dependency relations among objects. Third, the existing methods ignore the impacts of imbalance between hard and easy training pairs, resulting in suboptimal hash codes. To address these issues, we propose a novel deep hashing method, termed multi-label hashing for dependency relations among multiple objectives (DRMH). We first utilize an object detection network to extract object feature representations to avoid ignoring small object features and then fuse object visual features with position features and further capture dependency relations among objects using a self-attention mechanism. In addition, we design a weighted pairwise hash loss to solve the imbalance problem between hard and easy training pairs. Extensive experiments are conducted on multi-label datasets and zero-shot datasets, and the proposed DRMH outperforms many state-of-the-art hashing methods with respect to different evaluation metrics.

Hash Bit Selection With Reinforcement Learning for Image Retrieval

In recent years, binary hashing methods have been widely used in large-scale multimedia retrieval because of the low computational complexity and memory cost. Generally, better retrieval accuracy can be achieved with a longer hash code, which, however, may suffer redundancy. In this paper, we propose a novel hash bit selection method, called Hash Bit Selection with Reinforcement Learning (HBS-RL), which aims to adaptively select the most informative bits from the database binary codes. In our approach, the hash bit selection problem is firstly modeled as a Markov Decision Process (MDP), which is solved with reinforcement learning. HBS-RL learns a policy for bit selection, which effectively identifies the most informative bits by directly maximizing mean Average Precision (mAP) during training. Specially, given a generated bit pool, our HBS-RL can sequentially select bits with different code lengths with a very lightweight fully-connected policy network. The proposed method is evaluated on the MNIST, CIFAR-10, ImageNet and NUS-WIDE datasets, and the results show that it significantly improves the retrieval performance of the existing unsupervised and deep supervised hashing methods. It also outperforms the state-of-the-art bit selection methods. For convenience of repeating our results, we release our source code at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/xyez/HBS-RL</uri> .

Feature adaptive multi-view hash for image search

&lt;abstract&gt;&lt;p&gt;With the rapid development of network technology and small handheld devices, the amount of data has significantly increased and various kinds of data can be supplied to us at the same time. Recently, hashing technology has become popular in executing large-scale similarity search and image matching tasks. However, most of the prior hashing methods are mainly focused on the choice of the high-dimensional feature descriptor for learning effective hashing functions. In practice, real world image data collected from multiple scenes cannot be descriptive enough by using a single type of feature. Recently, several unsupervised multi-view hashing learning methods have been proposed based on matrix factorization, anchor graph and metric learning. However, large quantization error will be introduced via a sign function and the robustness of multi-view hashing is ignored. In this paper we present a novel feature adaptive multi-view hashing (FAMVH) method based on a robust multi-view quantization framework. The proposed method is evaluated on three large-scale benchmarks CIFAR-10, CIFAR-20 and Caltech-256 for approximate nearest neighbor search task. The experimental results show that our approach can achieve the best accuracy and efficiency in the three large-scale datasets.&lt;/p&gt;&lt;/abstract&gt;

Adaptive Marginalized Semantic Hashing for Unpaired Cross-Modal Retrieval

In recent years, Cross-Modal Hashing (CMH) has attracted much attention due to its fast query speed and efficient storage. Previous studies have achieved promising results for Cross-Modal Retrieval (CMR) by discovering discriminative hash codes and modality-specific hash functions. Nonetheless, most existing CMR works are subjected to some restrictions: 1) It is assumed that data of different modalities are fully paired, which is impractical in real applications due to sample missing and false data alignment, and 2) binary regression targets including the label matrix and binary codes are too rigid to effectively learn semantic-preserving hash codes and hash functions. To address these problems, this paper proposes an Adaptive Marginalized Semantic Hashing (AMSH) method which not only enhances the discrimination of latent representations and hash codes by adaptive margins, but can also be used for both paired and unpaired CMR. As a two-step method, in the first step, AMSH generates semantic-aware modality-specific latent representations with adaptively marginalized labels, thereby enlarging the distances between different classes, and exploiting the labels to preserve the inter-modal and intra-modal semantic similarities into latent representations and hash codes. In the second step, adaptive margin matrices are embedded into the hash codes, and enlarge the gaps between positive and negative bits, which improves the discrimination and robustness of hash functions. On this basis, AMSH generates similarity-preserving hash codes and robust hash functions without the strict one-to-one data correspondence requirement. Experiments are conducted on several benchmark datasets to demonstrate the superiority and flexibility of AMSH over some state-of-the-art CMR methods. The source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/LKYLKYZ/AMSH</uri> .

Performance Enhancement of Distributed System Using HDFS Federation and Sharding

In today's world, 2.5 exabytes of data are generated and processed by the IT industry and different organizations. Processing and managing such a massive volume of Big data is challenging. Hadoop(HDFS) is widely used framework for processing Bigdata. The limitation of HDFS is it can scale only the datanodes and not namenode where the metadata is managed. There is no scaling of name nodes in the present Hadoop architecture, so managing the metadata of the exabyte scale becomes an essential and challenging leading to tightly coupled block storage, namespace scalability, and performance bottleneck issues. The proposed approach dynamic federated metadata management (DFMM) architecture manages the metadata in the name node. DFMM configures multiple name nodes using HDFS federation technique, and metadata is partitioned into numerous name nodes using sharding technique. Sharding manages the metadata using locality-preserving hashing and consistent hashing methods. Locality-preserving hashing stores metadata in the most appropriate name node using the file's parent directory, and metadata is dynamically distributed among Name nodes to maintain load balancing. DFMM also maintains replicas for the high availability and reliability of data. Results showcase the superiority of the proposed DFMM architecture combined with the sharding technique is highly scalable and manages metadata efficiently, and can store up to 1 billion files with a namespace size of 400 GB, enhancing the throughput by 21% and is highly scalable as compared to existing techniques. The comparative study analysis of existing HDFS, and proposed DFMM architecture combined with sharding enhances the performance and manages the metadata more efficiently.

Label-Affinity Self-Adaptive Central Similarity Hashing for Image Retrieval

Due to the usage of global similarity, the hashing methods based on predefined hash centers have achieved more accurate retrieval results than the pairwise/triplet-based methods. Nevertheless, the fixed hash centers lack the perception of data distribution and are limited by the pre-determined Hadamard matrix, which consider neither the label semantic information nor the object scale size, resulting in sub-optimal retrieval performance and weak generalization ability. In this paper, we (1) adopt the label semantic information to generate self-adaptive hash centers and (2) propose the label-affinity coefficient ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">lac</i> ) that considers the scale size of each label/object appearing in the given image to calculate the real hash centroid for this image. Based on this, we propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Label-affinity Self-adaptive Central Similarity Hashing (LSCSH)</i> for image retrieval. LSCSH consists of a hash code generator module and a hash center adapter module. First, we obtain the label word vector ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , the word vector representation of each class label) via the Word2Vector technique to generate and update the hash centers that adapt to the distribution of both label word vectors and generated hash codes. Second, we learn <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">lac</i> to indicate the dominance of different labels corresponding to objects in each given image, which considers the unequal scales of each object (corresponding to a label) to calculate a more accurate hash centroid for each image. Last but not least, we design an asynchronous learning mechanism to enable each hash code and its corresponding hash centroid to adapt to each other dynamically. We conduct extensive experiments on 5 image datasets including CIFAR-10, ImageNet, VOC2012, MS-COCO and NUS-WIDE. The experimental results demonstrate that LSCSH can achieve the state-of-the-art visual retrieval performance on both single-label and multi-label image datasets. The code of this work is released at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/lzHZWZ/LSCSH_sourcecode.git</uri> .

Deep quantization network with visual-semantic alignment for zero-shot image retrieval

&lt;abstract&gt;&lt;p&gt;Approximate nearest neighbor (ANN) search has become an essential paradigm for large-scale image retrieval. Conventional ANN search requires the categories of query images to been seen in the training set. However, facing the rapid evolution of newly-emerging concepts on the web, it is too expensive to retrain the model via collecting labeled data with the new (unseen) concepts. Existing zero-shot hashing methods choose the semantic space or intermediate space as the embedding space, which ignore the inconsistency of visual space and semantic space and suffer from the hubness problem on the zero-shot image retrieval task. In this paper, we present an novel deep quantization network with visual-semantic alignment for efficient zero-shot image retrieval. Specifically, we adopt a multi-task architecture that is capable of $ 1) $ learning discriminative and polymeric image representations for facilitating the visual-semantic alignment; $ 2) $ learning discriminative semantic embeddings for knowledge transfer; and $ 3) $ learning compact binary codes for aligning the visual space and the semantic space. We compare the proposed method with several state-of-the-art methods on several benchmark datasets, and the experimental results validate the superiority of the proposed method.&lt;/p&gt;&lt;/abstract&gt;