Semantic-Driven Interpretable Deep Multi-Modal Hashing for Large-Scale Multimedia Retrieval

Multi-modal hashing fuses multi-modal features at both offline training and online query stage for compact binary hash learning. It has aroused extensive attention in research filed of efficient large-scale multimedia retrieval. However, existing methods adopt batch-based learning scheme or unsupervised learning paradigm. They cannot efficiently handle the very common online streaming multi-modal data (for batch-learning methods), or learn the hash codes suffering from limited discriminative capability and less flexibility for varied streaming data (for existing online multi-modal hashing methods). In this paper, we develop a supervised Flexible Online Multi-modal Hashing (FOMH) method to adaptively fuse heterogeneous modalities and flexibly learn the discriminative hash code for the newly coming data, even if part of the modalities is missing. Specifically, instead of adopting the fixed weights, the modalities weights in FOMH are automatically learned with the proposed flexible multi-modal binary projection to timely capture the variations of streaming samples. Further, we design an efficient asymmetric online supervised hashing strategy to enhance the discriminative capability of the hash codes, while avoiding the challenging symmetric semantic matrix decomposition and storage cost. Moreover, to support fast hash updating and avoid the propagation of binary quantization errors in online learning process, we propose to directly update the hash codes with an efficient discrete online optimization. Experiments on several public multimedia retrieval datasets validate the superiority of the proposed method from various aspects.

Due to the fast retrieval speed and low storage cost, cross-modal hashing has become the primary method for cross-modal retrieval. Since the emergence of deep cross-modal hashing methods, cross-modal retrieval significantly improved. However, the existing cross-modal hash retrieval methods still need to effectively utilize the dataset’s supervisory information and the lack of similarity expression ability. This means that the label information needs to be maximized, and the potential semantic relationship between two modalities cannot be fully explored, thus affecting the judgment of semantic similarity between two modalities. To address these problems, this paper proposes Tri-CMH, a cross-modal hash retrieval method with fused triples, which is an end-to-end modeling framework consisting of two parts: feature extraction and hash learning. Firstly, the multi-modal data are preprocessing into the form of triple groups. The data supervision matrix is constructed so that the samples with labels and their meanings are aggregated together. In contrast, the samples with labels and their opposite meanings are separated, thus avoiding the problem of the under-utilization of supervisory information in the data set and achieving the effect of efficiently utilizing the global supervisory information. Meanwhile, the loss function of the hash learning part is optimized by considering the Hamming distance loss, single-modality internal loss, cross-modality loss, and quantization loss to explicitly constrain semantically similar hash codes and semantically dissimilar hash codes and to improve the model’s ability to judge cross-modality semantic similarity. The method is trained and tested on the IAPR-TC12, MIRFLICKR-25K, and NUS-WIDE datasets, and the experimental evaluation criteria are mAP and PR curve, and the experimental results show the effectiveness and practicality of the method.

Many achievements have been made on learning to hash for uni-modal and cross-modal retrieval. However, it is still an unsolved problem that how to directly and efficiently learn discriminative discrete hash codes for the multimedia retrieval, where both query and database samples are represented with heterogeneous multi-modal features. With this motivation, we propose a Fast Discrete Collaborative Multi-modal Hashing (FDCMH) method in this paper. We first propose an efficient collaborative multi-modal mapping that first transforms heterogeneous multi-modal features into the unified factors to exploit the complementarity of multi-modal features and preserve the semantic correlations in multiple modalities with linear computation and space complexity. Such shared factors also bridge the heterogeneous modality gap and remove the inter-modality redundancy. Further, we develop an asymmetric hashing learning module to simultaneously correlate the learned hash codes with low-level data distribution and high-level semantics. In particular, this design could avoid the challenging symmetric semantic matrix factorization and $O(n^2)$ O ( n 2 ) memory cost ( $n$ n is the number of training samples). It can support both computation and memory efficient discrete hash optimization. Experiments on several public multimedia retrieval datasets demonstrate the superiority of the proposed approach compared with state-of-the-art hashing techniques, in terms of both model learning efficiency and retrieval accuracy.

Hashing could significantly accelerate large-scale image search by transforming the high-dimensional features into binary Hamming space, where efficient similarity search can be achieved with very fast Hamming distance computation and extremely low storage cost. As an important branch of hashing methods, multi-view hashing takes advantages of multiple features from different views for binary hash learning. However, existing multi-view hashing methods are either based on shallow models which fail to fully capture the intrinsic correlations of heterogeneous views, or unsupervised deep models which suffer from insufficient semantics and cannot effectively exploit the complementarity of view features. In this paper, we propose a novel Deep Collaborative Multi-view Hashing (DCMVH) method to deeply fuse multi-view features and learn multi-view hash codes collaboratively under a deep architecture. DCMVH is a new deep multi-view hash learning framework. It mainly consists of 1) multiple view-specific networks to extract hidden representations of different views, and 2) a fusion network to learn multi-view fused hash code. DCMVH associates different layers with instance-wise and pair-wise semantic labels respectively. In this way, the discriminative capability of representation layers can be progressively enhanced and meanwhile the complementarity of different view features can be exploited effectively. Finally, we develop a fast discrete hash optimization method based on augmented Lagrangian multiplier to efficiently solve the binary hash codes. Experiments on public multi-view image search datasets demonstrate our approach achieves substantial performance improvement over state-of-the-art methods.

Hashing can compress heterogeneous high-dimensional data into compact binary codes while preserving the similarity to facilitate efficient retrieval and storage, and thus hashing has recently received much attention from information retrieval researchers. Most of the existing hashing methods first predefine a fixed length (e.g., 32, 64, or 128 bit) for the hash codes before learning them with this fixed length. However, one sample can be represented by various hash codes with different lengths, and thus there must be some associations and relationships among these different hash codes because they represent the same sample. Therefore, harnessing these relationships will boost the performance of hashing methods. Inspired by this possibility, in this study, we propose a new model jointly multiple hash learning (JMH), which can learn hash codes with multiple lengths simultaneously. In the proposed JMH method, three types of information are used for hash learning, which come from hash codes with different lengths, the original features of the samples and label. In contrast to the existing hashing methods, JMH can learn hash codes with different lengths in one step. Users can select appropriate hash codes for their retrieval tasks according to the requirements in terms of accuracy and complexity. To the best of our knowledge, JMH is one of the first attempts to learn multi-length hash codes simultaneously. In addition, in the proposed model, discrete and closed-form solutions for variables can be obtained by cyclic coordinate descent, thereby making the proposed model much faster during training. Extensive experiments were performed based on three benchmark datasets and the results demonstrated the superior performance of the proposed method.

Multimodal hashing has attracted much interest for cross-modal similarity search on large-scale multimedia data sets because of its efficiency and effectiveness. Recently, supervised multimodal hashing, which tries to preserve the semantic information obtained from the labels of training data, has received considerable attention for its higher search accuracy compared with unsupervised multimodal hashing. Although these algorithms are promising, they are mainly designed to preserve pairwise similarities. When semantic labels of training data are given, the algorithms often transform the labels into pairwise similarities, which gives rise to the following problems: (1) constructing pairwise similarity matrix requires enormous storage space and a large amount of calculation, making these methods unscalable to large-scale data sets; (2) transforming labels into pairwise similarities loses the category information of the training data. Therefore, these methods do not enable the hash codes to preserve the discriminative information reflected by labels and, hence, the retrieval accuracies of these methods are affected. To address these challenges, this paper introduces a simple yet effective supervised multimodal hashing method, called label consistent matrix factorization hashing (LCMFH), which focuses on directly utilizing semantic labels to guide the hashing learning procedure. Considering that relevant data from different modalities have semantic correlations, LCMFH transforms heterogeneous data into latent semantic spaces in which multimodal data from the same category share the same representation. Therefore, hash codes quantified by the obtained representations are consistent with the semantic labels of the original data and, thus, can have more discriminative power for cross-modal similarity search tasks. Thorough experiments on standard databases show that the proposed algorithm outperforms several state-of-the-art methods.

Unsupervised cross-modal hashing has attracted considerable attention to support large-scale cross-modal retrieval. Although promising progresses have been made so far, existing methods still suffer from limited capability on excavating and preserving the intrinsic multi-modal semantics. In this paper, we propose a Correlation-Identity Reconstruction Hashing (CIRH) method to alleviate this challenging problem. We develop a new unsupervised deep cross-modal hash learning framework to model and preserve the heterogeneous multi-modal correlation semantics into both hash codes and functions, and simultaneously, we involve both the hash codes and functions with the descriptive identity semantics. Specifically, we construct a multi-modal collaborated graph to model the heterogeneous multi-modal correlations, and jointly perform the intra-modal and cross-modal semantic aggregation on homogeneous and heterogeneous graph networks to generate a multi-modal complementary representation with correlation reconstruction. Furthermore, an identity semantic reconstruction process is designed to involve the generated representation with identity semantics by reconstructing the input modality representations. Finally, we propose a correlation-identity consistent hash function learning strategy to transfer the modelled multi-modal semantics into the neural networks of modality-specific deep hash functions. Experiments demonstrate the superior performance of the proposed method on both retrieval accuracy and efficiency. We provide our source codes and experimental datasets at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/XizeWu/CIRH</uri>

Benefitting from the low storage cost and high retrieval efficiency, hash learning has become a widely used retrieval technology to approximate nearest neighbors. Within it, the cross-modal medical hashing has attracted an increasing attention in facilitating efficiently clinical decision. However, there are still two main challenges in weak multi-manifold structure perseveration across multiple modalities and weak discriminability of hash code. Specifically, existing cross-modal hashing methods focus on pairwise relations within two modalities, and ignore underlying multi-manifold structures across over 2 modalities. Then, there is little consideration about discriminability, i.e., any pair of hash codes should be different. In this paper, we propose a novel hashing method named multi-manifold deep discriminative cross-modal hashing (MDDCH) for large-scale medical image retrieval. The key point is multi-modal manifold similarity which integrates multiple sub-manifolds defined on heterogeneous data to preserve correlation among instances, and it can be measured by three-step connection on corresponding hetero-manifold. Then, we propose discriminative item to make each hash code encoded by hash functions be different, which improves discriminative performance of hash code. Besides, we introduce Gaussian-binary Restricted Boltzmann Machine to directly output hash codes without using any continuous relaxation. Experiments on three benchmark datasets (AIBL, Brain and SPLP) show that our proposed MDDCH achieves comparative performance to recent state-of-the-art hashing methods. Additionally, diagnostic evaluation from professional physicians shows that all the retrieved medical images describe the same object and illness as the queried image.

An efficient dual semantic preserving hashing for cross-modal 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.

The content-based remote sensing image retrieval (CBRSIR) has attracted increasing attention with the number of remote sensing (RS) images growing explosively. Benefiting from the strong capacity of the deep convolutional neural network (DCNN), the performance of CBRSIR has been improved in recent years. Although great successes have been obtained, learning the RS images' representative features and enhancing the retrieval efficiency for the large-scale CBRSIR tasks are still two challenging problems. In this article, we propose a new CBRSIR method named feature and hash (FAH) learning, which consists of a deep feature learning model (DFLM) and an adversarial hash learning model (AHLM). The DFLM aims at learning the RS images' dense features to guarantee the retrieval precision. In the DFLM, the DCNN and the proposed feature aggregation are integrated to capture the multiscale features. Then, the discrimination of the obtained features can be highlighted by the attention map in the developed attention branch. The AHLM maps the dense features onto the compact hash codes so that the retrieval efficiency can be improved. The AHLM contains a hash learning submodel and an adversarial regularization submodel. In particular, the hash learning submodel learns the real-valued hash codes that are similarity preserved by semantic supervisions. The adversarial regularization submodel regularizes the real-valued hash codes to learn the discrete uniform distribution with possible values 0 and 1. In this way, the hash codes are coding-balanced and the quantization errors are reduced. Encouraging experimental results counted on three public benchmark data sets demonstrate that our FAH can achieve competitive performance in the CBRSIR task compared with many existing hash learning methods.

Hashing has been widely used for large-scale remote sensing image retrieval due to its outstanding advantages in storage and search speed. Recently, deep hashing methods, which produce discriminative hash codes by building end-to-end deep convolutional networks, have shown promising results. However, training these networks requires numerous labeled images, which are scarce and expensive in remote sensing datasets. In order to solve this problem, we propose a deep unsupervised hashing method, namely deep contrastive self-supervised hashing (DCSH), which uses only unlabeled images to learn accurate hash codes. It eliminates the need for label annotation by maximizing the consistency of different views generated from the same image. More specifically, we assume that the hash codes generated from different views of the same image are similar, and those generated from different images are dissimilar. On the basis of the hypothesis, we can develop a novel loss function containing the temperature-scaled cross-entropy loss and the quantization loss to train the developed deep network end-to-end, resulting in hash codes with semantic similarity preserved. Our proposed network contains four parts. First, each image is transformed into two different views using data augmentation. After that, they are fed into an encoder with the same shared parameters to obtain deep discriminate features. Following this, a hash layer converts the high-dimensional image representations into compact binary codes. Lastly, a novel hash function is introduced to train the proposed network end-to-end and thus guide generated hash codes with semantic similarity. Extensive experiments on two popular benchmark datasets of the UC Merced Land Use Database and the Aerial Image Dataset have demonstrated that our DCSH has significant superiority in remote sensing image retrieval compared with state-of-the-art unsupervised hashing methods.

Hashing-based approximate nearest neighbor (ANN) search in huge databases has become popular due to its computational and memory efficiency. The popular hashing methods, e.g., Locality Sensitive Hashing and Spectral Hashing, construct hash functions based on random or principal projections. The resulting hashes are either not very accurate or are inefficient. Moreover, these methods are designed for a given metric similarity. On the contrary, semantic similarity is usually given in terms of pairwise labels of samples. There exist supervised hashing methods that can handle such semantic similarity, but they are prone to overfitting when labeled data are small or noisy. In this work, we propose a semi-supervised hashing (SSH) framework that minimizes empirical error over the labeled set and an information theoretic regularizer over both labeled and unlabeled sets. Based on this framework, we present three different semi-supervised hashing methods, including orthogonal hashing, nonorthogonal hashing, and sequential hashing. Particularly, the sequential hashing method generates robust codes in which each hash function is designed to correct the errors made by the previous ones. We further show that the sequential learning paradigm can be extended to unsupervised domains where no labeled pairs are available. Extensive experiments on four large datasets (up to 80 million samples) demonstrate the superior performance of the proposed SSH methods over state-of-the-art supervised and unsupervised hashing techniques.

Hashing has drawn more and more attention in image retrieval due to its high search speed and low storage cost. Traditional hashing methods project the high-dimensional hand-crafted visual features to compact binary codes by linear or non-linear hashing functions. Deep hashing methods, which integrate image representation learning and hash functions learning into a unified framework, have shown more superior performance. Most of existing supervised deep hashing methods mainly consider the semantic similarities among images by using pair-wise or triplet-wise constraints as supervision information. However, as a kind of crucial information, the rankings of the retrieval results, are neglected. Consequently, the produced hash codes may be suboptimal. In this paper, a new Deep Hashing with Top Similarity Preserving (DHTSP) method is proposed to optimize the quality of hash codes for image retrieval. Specifically, we utilize AlexNet to extract discriminative image representations directly from the raw image pixels and learn hash functions simultaneously. Then a top similarity preserving loss function is designed to preserve the similarity of returned images at the top of the ranking list. Experimental results on three benchmark datasets show that our proposed method outperforms most of state-of-the-art deep hashing methods and traditional hashing methods.

Similarity search plays an important role in many applications involving high-dimensional data. Due to the known dimensionality curse, the performance of most existing indexing structures degrades quickly as the feature dimensionality increases. Hashing methods, such as locality sensitive hashing (LSH) and its variants, have been widely used to achieve fast approximate similarity search by trading search quality for efficiency. However, most existing hashing methods make use of randomized algorithms to generate hash codes without considering the specific structural information in the data. In this paper, we propose a novel hashing method, namely, robust hashing with local models (RHLM), which learns a set of robust hash functions to map the high-dimensional data points into binary hash codes by effectively utilizing local structural information. In RHLM, for each individual data point in the training dataset, a local hashing model is learned and used to predict the hash codes of its neighboring data points. The local models from all the data points are globally aligned so that an optimal hash code can be assigned to each data point. After obtaining the hash codes of all the training data points, we design a robust method by employing l2,1 -norm minimization on the loss function to learn effective hash functions, which are then used to map each database point into its hash code. Given a query data point, the search process first maps it into the query hash code by the hash functions and then explores the buckets, which have similar hash codes to the query hash code. Extensive experimental results conducted on real-life datasets show that the proposed RHLM outperforms the state-of-the-art methods in terms of search quality and efficiency.