Visual Feature Space Research Articles

Visual Saliency Map from Tensor Analysis

Modeling visual saliency map of an image provides important information for image semantic understanding in many applications. Most existing computational visual saliency models follow a bottom-up framework that generates independent saliency map in each selected visual feature space and combines them in a proper way. Two big challenges to be addressed explicitly in these methods are (1) which features should be extracted for all pixels of the input image and (2) how to dynamically determine importance of the saliency map generated in each feature space. In order to address these problems, we present a novel saliency map computational model based on tensor decomposition and reconstruction. Tensor representation and analysis not only explicitly represent image's color values but also imply two important relationships inherent to color image. One is reflecting spatial correlations between pixels and the other one is representing interplay between color channels. Therefore, saliency map generator based on the proposed model can adaptively find the most suitable features and their combinational coefficients for each pixel. Experiments on a synthetic image set and a real image set show that our method is superior or comparable to other prevailing saliency map models.

Zero-shot learning via a specific rank-controlled semantic autoencoder

Existing embedding zero-shot learning models usually learn a projection function from the visual feature space to the semantic embedding space, e.g. attribute space or word vector space. However, the projection learned based on seen samples may not generalize well to unseen classes, which is known as the projection domain shift problem in ZSL. To address this issue, we propose a method named Low-rank Semantic Autoencoder (LSA) to consider the low-rank structure of seen samples to maintain the sparse feature of reconstruction error, which can further improve zero-shot learning capability. Moreover, to obtain a more robust projection for unseen classes, we propose a Specific Rank-controlled Semantic Autoencoder (SRSA) to accurately control of the projection’s rank. Extensive experiments on six benchmarks demonstrate the superiority of the proposed models over most existing embedding ZSL models under the standard zero-shot setting and the more realistic generalized zero-shot setting.

A General Visual-Impedance Framework for Effectively Combining Vision and Force Sensing in Feature Space

Robotic systems are increasingly used to work in dynamic and/or unstructured environments and to operate with a high degree of safety and autonomy. Consequently, they are often equipped with external sensors capable of perceiving the environment (e.g., cameras) and the contacts that may arise (e.g. force/torque sensors). This letter proposes a general framework for combining force and visual information in the visual feature space. By leveraging recent results on the derivation of visual servo dynamics, we generalize the treatment regardless of the visual features chosen. Vision and force sensing are coupled in the feature space, avoiding both the convergence to a local minimum and the arising of inconsistencies at the actuation level. Any task space direction is simultaneously controlled by both vision and force. Compliance against interaction forces is achieved in feature space along the features defining the visual task. Experiments on a real platform are carried out to evaluate the effectiveness of the proposed framework.

Learning Multipart Attention Neural Network for Zero-Shot Classification

Zero-shot learning (ZSL) models typically learn a cross-modal mapping between the visual feature space and the semantic embedding space. Despite promising performance achieved by existing methods, they usually take visual features from the whole image as the main proposed inputs, while pay little attention to image regions which are relevant to human’s visual response to the whole image. In this article, we propose a neural network-based ZSL model which incorporates an attention mechanism to discover the discriminative parts for each image. The proposed model allows us to automatically generate attention maps for visual parts, which provides a flexible way of encoding the salient visual aspects to distinguish the categories. Moreover, we introduce a simple yet effective objective function to exploit the pairwise label information between images and classes, resulting in substantial performance improvement. When multiple semantic spaces are available, a multiple-attention scheme is provided to fuse different semantic spaces, which helps to achieve further improvement in performance. On the widely used CUB-2010-2011 data set for fine-grained image classification, we demonstrate the advantages of using attention mechanism and semantic parts in our model for ZSL. Comprehensive experimental results show that our proposed approach achieves superior performance than the state-of-the-art methods.

The image annotation algorithm using convolutional features from intermediate layer of deep learning

The automatic image annotation is an effective computer operation that predicts the annotation of an unknown image by automatically learning potential relationships between the semantic concept space and the visual feature space in the annotation image dataset. Usually, the auto-labeling image includes the processing: learning processing and labeling processing. Existing image annotation methods that employ convolutional features of deep learning methods have a number of limitations, including complex training and high space/time expenses associated with the image annotation procedure. Accordingly, this paper proposes an innovative method in which the visual features of the image are presented by the intermediate layer features of deep learning, while semantic concepts are represented by mean vectors of positive samples. Firstly, the convolutional result is directly output in the form of low-level visual features through the mid-level of the pre-trained deep learning model, with the image being represented by sparse coding. Secondly, the positive mean vector method is used to construct visual feature vectors for each text vocabulary item, so that a visual feature vector database is created. Finally, the visual feature vector similarity between the testing image and all text vocabulary is calculated, and the vocabulary with the largest similarity used for annotation. Experiments on the datasets demonstrate the effectiveness of the proposed method; in terms of F1 score, the proposed method’s performance on the Corel5k dataset and IAPR TC-12 dataset is superior to that of MBRM, JEC-AF, JEC-DF, and 2PKNN with end-to-end deep features.

Bidirectional generative transductive zero-shot learning

Most zero-shot learning (ZSL) methods aim to learn a mapping from visual feature space to semantic feature space or from both visual and semantic feature spaces to a common joint space and align them. However, in these methods the visual and semantic information are not utilized sufficiently and the useless information is not excluded. Moreover, there exists a strong bias problem that the instances from unseen classes always tend to be predicted as some seen classes in most ZSL methods. In this paper, combining the advantages of generative adversarial networks (GANs), a method based on bidirectional projections between the visual and semantic feature spaces is proposed. GANs are used to perform bidirectional generations and alignments between the visual and semantic features. In addition, cycle mapping structure ensures that the important information are kept in the alignments. Furthermore, in order to better solve the bias problem, pseudo-labels are generated for unseen instances and the model is adjusted along with them iteratively. We conduct extensive experiments at traditional ZSL and generalized ZSL settings, respectively. Experiment results confirm that our method achieves the state-of-the-art performances on the popular datasets AWA2, aPY and SUN.

Inductive Structure Consistent Hashing via Flexible Semantic Calibration.

Semantic-preserving hashing establishes efficient multimedia retrieval by transferring knowledge from original data to hash codes so that the latter can preserve the underlying visual and semantic similarities. However, it becomes a crucial bottleneck: how to effectively bridge the trilateral domain gaps (i.e., the visual, semantic, and hashing spaces) to further improve the retrieval accuracy. In this article, we propose an inductive structure consistent hashing (ISCH) method, which can interactively coordinate the semantic correlations between the visual feature space, the binary class space, and the discrete hashing space. Specifically, an inductive semantic space is formulated by a simple multilayer stacking class-encoder, which transforms the naive class information into flexible semantic embeddings. Meanwhile, we design a semantic dictionary learning model to facilitate the bilateral visual-semantic bridging and guide the class-encoder toward reliable semantics, which could well alleviate the visual-semantic bias problem. In particular, the visual descriptors and respective semantic class representations are regularized with a coinciding alignment module. In order to generate privileged hash codes, we further explore semantic and prototype binary code learning to jointly quantify the semantic and latent visual representations into unified discrete hash codes. Moreover, an efficient optimization algorithm is developed to address the resulting discrete programming problem. Comprehensive experiments conducted on four large-scale data sets, i.e., CIFAR-10, NUSWIDE, ImageNet, and MSCOCO, demonstrate the superiority of our method over the state-of-the-art alternatives against different evaluation protocols.

A Brain-Media Deep Framework Towards Seeing Imaginations Inside Brains

While current research on multimedia is essentially dealing with the information derived from our observations of the world, internal activities inside human brains, such as imaginations and memories of past events etc., could become a brand new concept of multimedia, for which we coin as “brain-media”. In this paper, we pioneer this idea by directly applying natural images to stimulate human brains and then collect the corresponding electroencephalogram (EEG) sequences to drive a deep framework to learn and visualize the corresponding brain activities. By examining the relevance between the visualized image and the stimulation image, we are able to assess the performance of our proposed deep framework in terms of not only the quality of such visualization but also the feasibility of introducing the new concept of “brain-media”. To ensure that our explorative research is meaningful, we introduce a dually conditioned learning mechanism in the proposed deep framework. One condition is analyzing EEG sequences through deep learning to extract a more compact and class-dependent brain features via exploiting those unique characteristics of human brains such as hemispheric lateralization and biological neurons myelination (neurons importance), and the other is to analyze the content of images via computing approaches and extract representative visual features to exploit artificial intelligence in assisting our automated analysis of brain activities and their visualizations. By combining the brain feature space with the associated visual feature space of those images that are candidates of the stimuli, we are able to generate a combined-conditional space to support the proposed dual-conditioned and lateralization-supported GAN framework. Extensive experiments carried out illustrate that our proposed deep framework significantly outperforms the existing relevant work, indicating that our proposed does provide a good potential for further research upon the introduced concept of “brain-media”, a new member for the big family of multimedia. To encourage more research along this direction, we make our source codes publicly available for downloading at GitHub. 1 1 https://github.com/aneeg/LS-GAN .

A Novel Perspective to Zero-Shot Learning: Towards an Alignment of Manifold Structures via Semantic Feature Expansion

Zero-shot learning aims at recognizing unseen classes (no training example) with knowledge transferred from seen classes. This is typically achieved by exploiting a semantic feature space shared by both seen and unseen classes, i.e., attribute or word vector, as the bridge. One common practice in zero-shot learning is to train a projection between the visual and semantic feature spaces with labeled seen classes examples. When inferring, this learned projection is applied to unseen classes and recognizes the class labels by some metrics. However, the visual and semantic feature spaces are mutually independent and have quite different manifold structures. Under such a paradigm, most existing methods easily suffer from the domain shift problem and weaken the performance of zero-shot recognition. To address this issue, we propose a novel model called AMS-SFE. It considers the alignment of manifold structures by semantic feature expansion. Specifically, we build upon an autoencoder-based model to expand the semantic features from the visual inputs. Additionally, the expansion is jointly guided by an embedded manifold extracted from the visual feature space of the data. Our model is the first attempt to align both feature spaces by expanding semantic features and derives two benefits: first, we expand some auxiliary features that enhance the semantic feature space; second and more importantly, we implicitly align the manifold structures between the visual and semantic feature spaces; thus, the projection can be better trained and mitigate the domain shift problem. Extensive experiments show significant performance improvement, which verifies the effectiveness of our model.

A Variational Autoencoder with Deep Embedding Model for Generalized Zero-Shot Learning

Generalized zero-shot learning (GZSL) is a challenging task that aims to recognize not only unseen classes unavailable during training, but also seen classes used at training stage. It is achieved by transferring knowledge from seen classes to unseen classes via a shared semantic space (e.g. attribute space). Most existing GZSL methods usually learn a cross-modal mapping between the visual feature space and the semantic space. However, the mapping model learned only from the seen classes will produce an inherent bias when used in the unseen classes. In order to tackle such a problem, this paper integrates a deep embedding network (DE) and a modified variational autoencoder (VAE) into a novel model (DE-VAE) to learn a latent space shared by both image features and class embeddings. Specifically, the proposed model firstly employs DE to learn the mapping from the semantic space to the visual feature space, and then utilizes VAE to transform both original visual features and the features obtained by the mapping into latent features. Finally, the latent features are used to train a softmax classifier. Extensive experiments on four GZSL benchmark datasets show that the proposed model significantly outperforms the state of the arts.

Re-ranking image-text matching by adaptive metric fusion

Image-text matching has drawn much attention recently with the rapid growth of multi-modal data. Many effective approaches have been proposed to solve this challenging problem, but limited effort has been devoted to re-ranking methods. Compared with the uni-modal re-ranking methods, modality heterogeneity is the major difficulty when designing a re-ranking method in the cross-modal field, which mainly lies in two aspects of different visual and textual feature spaces and different distributions in inverse directions. In this paper, we propose a heuristic re-ranking method called Adaptive Metric Fusion (AMF) for image-text matching. The method can obtain a better metric by adaptively fusing metrics based on two modules: 1) Cross-modal Reciprocal Encoding, which considers ranks in inverse directions to comprehensively evaluate a metric. The sentence retrieval and image retrieval have different distribution characteristics and galleries in different modalities, thus it is necessary to exploit them simultaneously for appropriate metric fusion. 2) Query Replacement Gap, which quantifies the gap between cross-modal and uni-modal similarities to alleviate the influence of different visual and textual feature spaces on the fused metric. The proposed re-ranking method can be implemented in an unsupervised way without requiring any human interaction or annotated data, and can be easily applied to any initial ranking result. Extensive experiments and analysis validate the effectiveness of our method on the large-scale MS-COCO and Flickr30K datasets.

Anomaly Detection Using Local Kernel Density Estimation and Context-Based Regression

Current local density-based anomaly detection methods are limited in that the local density estimation and the neighborhood density estimation are not accurate enough for complex and large databases, and the detection performance depends on the size parameter of the neighborhood. In this paper, we propose a new kernel function to estimate samples’ local densities and propose a weighted neighborhood density estimation to increase the robustness to changes in the neighborhood size. We further propose a local kernel regression estimator and a hierarchical strategy for combining information from the multiple scale neighborhoods to refine anomaly factors of samples. We apply our general anomaly detection method to image saliency detection by regarding salient pixels in objects as anomalies to the background regions. Local density estimation in the visual feature space and kernel-based saliency score propagation in the image enable the assignment of similar saliency values to homogenous object regions. Experimental results on several benchmark datasets demonstrate that our anomaly detection methods overall outperform several state-of-art anomaly detection methods. The effectiveness of our image saliency detection method is validated by comparison with several state-of-art saliency detection methods.

Cooperative Coupled Generative Networks for Generalized Zero-Shot Learning

Compared with zero-shot learning (ZSL), the generalized zero-shot learning (GZSL) is more challenging since its test samples are taken from both seen and unseen classes. Most previous mapping-based methods perform well on ZSL, while their performance degrades on GZSL. To solve this problem, inspired by the ensemble learning, this paper proposes a model with cooperative coupled generative networks (CCGN). Firstly, to alleviate the hubness problem, the reverse visual feature space is taken as the embedding space, with the mapping achieved by a visual feature center generation network. To learn a proper visual representation of each class, we propose a coupled of generative networks, which cooperate with each other to synthesize a visual feature center template of the class. Secondly, to improve the generative ability of the coupled networks, we further employ a deeper network to generate. Meanwhile, to alleviate loss semantic information problem caused by multiple network layers, a residual module is employed. Thirdly, to mitigate overfitting and to increase scalability, an adversarial network is introduced to discriminate the generation of visual feature centers. Finally, a reconstruction network, which reverses the generation process, is employed to restrict the structural correlation between the generated visual feature center and the original semantic representation of each class. Extensive experiments on five benchmark datasets (AWA1, AWA2, CUB, SUN, APY) demonstrate that the proposed algorithm yields satisfactory results, as compared with the state-of-the-art methods.

Cross-domain mapping learning for transductive zero-shot learning

Zero-shot learning (ZSL) aims to learn a projection function from a visual feature space to a semantic embedding space or reverse. The main challenge of ZSL is the domain shift problem where the unseen test data has a large gap with the seen training data. Transductive ZSL based methods alleviate this problem by learning from both labeled data and unlabeled data to capture their common semantic information. In this paper, we propose a framework to learn a robust cross-domain mapping for transductive ZSL with an extremely efficient algorithm for model optimization. Combining with a deep model, we formulate the cross-domain mapping as a general loss function that optimizes both the projection function and discriminative visual features simultaneously in an end-to-end manner. Extensive experiments on five benchmark datasets show that the proposed Cross-Domain Mapping (CDM) model outperforms the state-of-the-art.

Large Margin Nearest Neighbor Classification With Privileged Information for Biometric Applications

In this paper, a new metric learning algorithm is proposed to improve face verification and person re-identification in RGB images by learning from RGB and Depth (RGB-D) training images. We address this problem by formulating it as a Learning Using Privileged Information problem, in which the additional depth images associated with the RGB training images are not available for the testing process. Based on the large margin nearest neighbor (LMNN) classification framework, we propose an effective metric learning method called large margin nearest neighbor classification with privileged information (LMNN+) by incorporating depth information to improve decision function learning in the training process. Specifically, two distance metrics based on visual features as well as depth features are jointly learned by minimizing the triplet loss in which the within-class difference is minimized, while the between-class difference is maximized. The distances in the depth feature space can be utilized to guide the training process in the visual feature space. In addition, we propose an efficient optimization method which can handle billions of constraints in the optimization problem of LMNN+. The comprehensive experiments on the EUROCOM data set, the CurtinFaces data set as well as the BIWI RGBD-ID data set demonstrate the effectiveness of our algorithm for face verification and person re-identification by leveraging privileged information.

Fusion by synthesizing: A multi-view deep neural network for zero-shot recognition

Zero-shot learning (ZSL) aims to recognize objects without seeing any visual instances by learning knowledge transfer between seen and unseen classes. Attributes, which denote high-level visual entities or visual characteristics, have been widely utilized as intermediate embedding space for knowledge transfer in a majority of existing ZSL approaches and have shown impressive performance. Attribute based ZSL approaches, which introduce an intermediate embedding space of attributes for knowledge transfer, have shown impressive performance. However, providing attribute annotations for unseen classes at test time is time-consuming and labor-intensive. Besides, directly using attributes as intermediation (embedding space) for knowledge transfer and zero-shot prediction inevitably leads to the projection domain shift and hubness problems. In this paper, we propose a novel multi-view deep neural network, termed Fusion by Synthesis (FS), which leverages word embeddings of classes as complementary for attributes and performs zero-shot prediction by fusing the word embeddings of unseen classes and the synthesized attributes in the visual feature space. Specifically, in the training phase, by considering the visual features, attributes and word embeddings as three different views of visual instances, FS allocates each view with a denoising auto-encoder to simultaneously ensures robust view-specific reconstructions and cross-view synthesizing, while preserving the discrimination of class labels. During testing, FS can synthesize the absent attributes for unseen classes and fuse them with word embeddings to the visual feature space to perform zero-shot prediction. Besides, FS is flexible to learn with partial views, where either attribute view or word embedding view is missing during training. Moreover, FS is flexible to synthesize either the missing view of attributes or word embeddings from the provided view. Extensive experiments on six benchmark datasets on both image classification and action recognition show that FS is advantaged to fuse multi-view data by synthesis and achieves superior performance compared with the state-of-the-art ZSL methods.

Unsupervised deep triplet hashing with pseudo triplets for scalable image retrieval

Deep learning based hashing methods have been proven to be effective in the field of image retrieval recently. Among them, most high-performance methods are supervised frameworks, which require annotated labels by humans. Considering the difficulty of labeling large-scale image datasets, unsupervised methods, which just need images themselves for training, are more suitable for practical applications. However, how to improve the discriminative ability of hash codes generated by unsupervised models still remains as a challenging problem. In this paper, we present a novel deep framework called Unsupervised Deep Triplet Hashing (UDTH) for scalable image retrieval. UDTH builds pseudo triplets based on the neighborhood structure in the high-dimensional visual feature space, and then solves two problems through the proposed objective function: 1) Triplet network is utilized to maximize the distance between different classes of binary representation; 2) Autoencoder and Binary quantization are exploited to learn hash codes which maintain the structural information of original samples. Extensive experiments on the datasets of CIFAR-10, NUS-WIDE and MIRFLICKR-25K are conducted, and the results show that our proposed UDTH is superior to the state-of-the-art methods.

Zero-Shot Learning via Category-Specific Visual-Semantic Mapping and Label Refinement.

Zero-Shot Learning (ZSL) aims to classify a test instance from an unseen category based on the training instances from seen categories, in which the gap between seen categories and unseen categories is generally bridged via visual-semantic mapping between the low-level visual feature space and the intermediate semantic space. However, the visual-semantic mapping (i.e., projection) learnt based on seen categories may not generalize well to unseen categories, which is known as the projection domain shift in ZSL. To address this projection domain shift issue, we propose a method named Adaptive Embedding ZSL (AEZSL) to learn an adaptive visual-semantic mapping for each unseen category, followed by progressive label refinement. Moreover, to avoid learning visual-semantic mapping for each unseen category in the large-scale classification task, we additionally propose a deep adaptive embedding model named Deep AEZSL (DAEZSL) sharing the similar idea (i.e., visual-semantic mapping should be category-specific and related to the semantic space) with AEZSL, which only needs to be trained once, but can be applied to arbitrary number of unseen categories. Extensive experiments demonstrate that our proposed methods achieve the state-of-theart results for image classification on three small-scale benchmark datasets and one large-scale benchmark dataset.

Zero-Shot Learning via Low-Rank-Representation Based Manifold Regularization

In this letter, we propose a novel low-rank-represen-tation (LRR) based manifold-regularization approach for zero-shot learning (ZSL). Most existing regularization-based ZSL approaches perform the alignment between visual feature space and semantic space based on the affinity matrix constructed from the test instances. The affinity matrix plays a significant role in exploiting the manifold structures of visual feature space, hence we propose to use the LRR to guide the affinity-matrix construction by exploring the subspace structures of data. Considering the locality and similarity information among data, we incorporate a Laplacian regularization term to the LRR framework to ensure that the learned affinity matrix can capture the local geometric structures in data. We also explicitly impose the nonnegative sparse constraint on the affinity matrix to facilitate the learning of local manifold structures. Moreover, we use an effective manifold-regularization methodology to learn discriminative semantic representations of test instances, leading to significant improvements in classification performance over the unseen classes. Extensive experiments on three benchmark datasets demonstrate that the proposed approach outperforms the state of the arts.

Generative Zero-Shot Learning via Low-Rank Embedded Semantic Dictionary.

Zero-shot learning for visual recognition, which approaches identifying unseen categories through a shared visual-semantic function learned on the seen categories and is expected to well adapt to unseen categories, has received considerable research attention most recently. However, the semantic gap between discriminant visual features and their underlying semantics is still the biggest obstacle, because there usually exists domain disparity across the seen and unseen classes. To deal with this challenge, we design two-stage generative adversarial networks to enhance the generalizability of semantic dictionary through low-rank embedding for zero-shot learning. In detail, we formulate a novel framework to simultaneously seek a two-stage generative model and a semantic dictionary to connect visual features with their semantics under a low-rank embedding. Our first-stage generative model is able to augment more semantic features for the unseen classes, which are then used to generate more discriminant visual features in the second stage, to expand the seen visual feature space. Therefore, we will be able to seek a better semantic dictionary to constitute the latent basis for the unseen classes based on the augmented semantic and visual data. Finally, our approach could capture a variety of visual characteristics from seen classes that are "ready-to-use" for new classes. Extensive experiments on four zero-shot benchmarks demonstrate that our proposed algorithm outperforms the state-of-the-art zero-shot algorithms.