Search

The image classification precision is vastly enhanced with the growing complexity of convolutional neural network (CNN) structures. However, the uneven visual separability between categories leads to various difficulties in classification. The hierarchical structure of categories can be leveraged to deal with it, but a few CNNs pay attention to the character of data. Besides, a network model with a hierarchical structure is promising to extract more specific features from the data than current CNNs, since, for the latter, all categories have the same fixed number of layers for feed-forward computation. In this paper, we propose to use category hierarchies to integrate ResNet-style modules to form a hierarchical network model in a top-down manner. To extract abundant discriminative features and improve the computation efficiency, we adopt residual block selection based on coarse categories to allocate different computation paths. Each residual block works as a switch to determine the JUMP or JOIN mode for an individual coarse category. Interestingly, since some categories need less feed-forward computation than others by jumping layers, the average inference time cost is reduced. Extensive experiments show that our hierarchical network achieves higher prediction accuracy with similar FLOPs on CIFAR-10 and CIFAR-100, SVHM, and Tiny-ImageNet datasets compared to original residual networks and other existing selection inference methods. • The hierarchy of classes is used to form a hierarchical network. • The selective computation based on coarse category is proposed for the first time. • The proposed network model helps inference acceleration and accuracy improvement. [ABSTRACT FROM AUTHOR]

• First paper to summarize existing adversarial examples attacks in the object detection field. • We innovatively classify the attacking models by the characteristic of object detection tasks which can be treated as a multitasking process. • The comparison between the two main processes of attacking object detection is generalized. • Analyzed different results for attacking regression and positioning processes in object detection tasks. Deep learning plays a critical role in the applications of artificial intelligence. The trend of processing images or videos as input data and pursuing execution efficiency in practical applications is unstoppable. However, the vulnerability due to the complex structure of deep networks makes it at risk of attacks. Object detection, as the significant product impacted by the deep learning frame, corresponds to this weakness implicated by its multiple-tasks property. Besides, these applications involving object detection techniques are integrated deeply into our lives, potentially leading to unimaginable loss. Adversarial example attacks, as the mainstream attack method, provide an efficacious and comprehensible idea to generate perturbation. In this survey, we review the existing adversarial example attacks in object detection tasks and inductively discuss the similarity and differences among these approaches. Finally, we construct this survey for discussing the attacks in the object detection field and point out the possible direction for adversarial defenses in future studies. [ABSTRACT FROM AUTHOR]

As a powerful machine learning method, deep learning has attracted the attention of numerous researchers. While exploring a high-performance neural network model, the floating-point operations of a neural network model are also increasing. In recent years, many researchers have noticed that efficiency is also one of important indicators to measure the property of neural network models. Obviously, the efficient neural network model is more helpful to deploy on mobile and embedded devices. Therefore, the efficient neural network model becomes a hot research spot. In this paper, we review the methods related to the structural design of efficient convolution neural networks in recent years. According to the characteristics of these methods, we divide them into three kinds of methods: model pruning, efficient architecture, and neural architecture search. Detailed analyses of each method are presented to demonstrate their advantages and disadvantages. Then, we comprehensively compare them in detail and propose many suggestions about the design of the efficient convolution neural network model structure. Inspired by these suggestions, we built a new efficient neural network model, SharedNet. And the SharedNet obtains the best accuracy of manually-designed efficient CNN models on the ImageNet dataset. [ABSTRACT FROM AUTHOR]

Collaborative representation-based (CR) methods have become prevalent for pattern classification tasks, achieving formidable performance. Theoretically, we expect the learned class-specific representation of the correct class to be discriminative against others, with the representation of the correct class contributing dominantly in CR. However, most existing CR methods focus on improving discrimination while having a limited impact on enhancing the representation contribution of the correct category. In this work, we propose a novel CR approach for image classification called the elastic competitive and discriminative collaborative representation-based classifier (ECDCRC) to simultaneously strengthen representation contribution and discrimination of the correct class. The ECDCRC objective function penalizes two key terms by fully incorporating label information. The competitive term integrates the nearest subspace representation with corresponding elastic factors into the model, allowing each class to have varying competition intensities based on similarity with the query sample. This enhances the representation contribution of the correct class in CR. To further improve discrimination, the discriminative term introduces an elastic factor as a weight in the model to represent the gap between the query sample and the representation of each class. Moreover, instead of focusing on representation coefficients, the designed ECDCRC weights associated with representation components directly relate to the representation of each class, enabling more direct and precise discrimination improvement. Concurrently, sparsity is also enhanced through the two terms, further boosting model performance. Additionally, we propose a robust ECDCRC (R-ECDCRC) to handle image classification with noise. Extensive experiments on seven public databases demonstrate the proposed method's superior performance over related state-of-the-art CR methods. • An elastic competitive and discriminative representation method is proposed. • Analyses of the discrimination, the representation contribution, and the sparsity. • Multiple experiments underscore the method's performance. [ABSTRACT FROM AUTHOR]

Symmetry is an important property of human faces. Many researchers exploit symmetry to improve face recognition. In this paper, we explore the symmetry of faces under the framework of PCA (Principal Component Analysis). Unlike previous studies that manipulate facial symmetry by averaging the two halves, we argue that well-estimated symmetrical faces lie in a low-dimensional subspace. Inspired by this, a PCA-like model, referred to as Mirror PCA, is proposed to extract low-rank symmetrical features of faces, as well as their principal directions. Although the optimization problem is non-convex, a suboptimal solution could be found under existing optimization frameworks such as ADMM, PGD, etc. A number of experiments are conducted to verify the effectiveness and robustness of our method. [ABSTRACT FROM AUTHOR]

Artificial intelligence and its derivative technologies are not only playing a role in the fields of medicine, economy, policing, transportation, and natural science computing today but also in future industries such as electric vehicles and meta-universe. However, because of the black-box nature of most common artificial neural networks (DNNs), there needs to be more understanding of what is happening behind these astounding performances. The abstract reasoning process in the networks raises concerns about the security of AI systems. Because of this, more and more researchers are turning their attention to the explainability of black-box neural networks to find some attributions that explain the reasoning performed by the networks by studying the black-box nature of the networks. Deep neural networks and their explainability act as a mutually beneficial symbiosis, facilitating each other's development. This survey reviews the deep network explainable methods applicable for the field of computing vision proposed within the last decade and categorizes these methods in terms of their starting point to explain deep neural networks. Focusing on their intrinsics, we review the methods that are applicable for object classification, object detection, and looked forward that of object tracking. In each cluster, we show that the methods share some similarities but have their own highlight. Furthermore, we shed light on the future development of the field of explainable AI. • Up-to-date explainable methods. • Explainable classifications and object detections. • Pros and cons and visual comparisons. • Demonstrations of most explainable methods reviewed. [ABSTRACT FROM AUTHOR]

Sparse representation based classification (SRC) has become a popular methodology in face recognition in recent years. One widely used manner is to enforce minimum l 1 -norm on coding coefficient vector, which is considered as an unsupervised sparsity constraint and usually requires high computational cost. On the other hand, supervised sparsity representation based method (SSR) realizes sparse representation classification with higher efficiency by multiple phases of representing a probe. Nevertheless, since previous SSR methods only deal with Gaussian noise, they cannot satisfy empirical face recognition application which faces wide variations. In this paper, we propose a robust supervised sparse representation (RSSR) model, which uses two-phase of robust representation to compute a sparse coding vector. Huber loss is employed as the fidelity term in the linear representation, which improves the competitiveness of correct class in the first phase. Then training samples with weak competitiveness are removed by supervised way. In the second phase, the competitiveness of correct class is further boosted by Huber loss. We compare the RSSR with other state-of-the-art methods under different conditions, including illumination variations, gesture changes, expressions, corruptions, and occlusions. Comprehensive experiments on four open databases demonstrate the robustness of RSSR and competitive performance is obtained in dealing with face images with occlusion or not. [ABSTRACT FROM AUTHOR]

Principal component analysis (PCA) is a widely used tool for dimensionality reduction and feature extraction in the field of computer vision. Traditional PCA is sensitive to outliers which are common in empirical applications. Therefore, in recent years, massive efforts have been made to improve the robustness of PCA. However, many emerging PCA variants developed in the direction have some weaknesses. First, few of them pay attention to the 2D structure of error matrix. Second, to estimate data mean from sample set with outliers by averaging is usually biased. Third, if some elements of a sample are disturbed, to extract principal components (PCs) by directly projecting data with transformation matrix causes incorrect mapping of sample to its genuine location in low-dimensional feature subspace. To alleviate these problems, we present a novel robust method, called nuclear norm-based on PCA (N-PCA) to take full advantage of the structure information of error image. Meanwhile, it is developed under a novel unified framework of PCA to remedy the bias of computing data mean and the low-dimensional representation of a sample both of which are treated as unknown variables in a single model together with projection matrix. To solve N-PCA, we propose an iterative algorithm, which has a closed-form solution in each iteration. Experimental results on several open databases demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Robust sparse representation is a well-known method in computer vision. Many sparse representation models have been proposed and perform well in face recognition. Most of them use transformed images of one dimensional vector, and such implementation ignores structural information between features. To make use of this structural information, this paper presents a novel model for face recognition. Unlike traditional sparse regression measuring differences between test sample and predicted response by vector norm, our model uses matrix norm, l 2,1 , to calculate residual. Instead of treating each pixel independently, the residual of a pixel is connected with all other pixels from the same line and column by means of double direction l 2,1 -norm. Moreover, we incorporate the specific class information into the representation model by introducing group-sparse regularizer. And then, we use the alternating direction method of multipliers approach to optimize proposed model. The extensive experiments have been conducted to evaluate the proposed methods by comparing with many sparse representation based techniques. [ABSTRACT FROM AUTHOR]

• A new matrix-based sparse representation classi fication approach is pro-posed. • Block-norm depresses the occlusion parts of an image and utilizes the rest parts. • A self-adaptive threshold is employed to restrict large residual error. In this paper, we consider the robust face recognition problem via matrix regression-based classification with block-norm(MRC-block). Specifically, we propose an inner-class sparse representation classification approach in which images are expressed as matrices instead of vectors and adopted to encode more discriminative information than other regression-based methods. In the regression step, a block-norm based matrix regression model is proposed, which can efficiently depress the effect of occlusion in probe images. Accordingly, an efficient algorithm is derived to optimize the proposed objective function. Comprehensive experiments on representative datasets demonstrate that MRC-block is a robust discriminative classifier which significantly improves the performance and efficiency compared with the state-of-the-art methods in dealing with face image with occlusion. [ABSTRACT FROM AUTHOR]

Recently linear representation provides an effective way for robust face recognition. However, the existing linear representation methods cannot make an adaptive adjustment in responding to the variations on facial image, so the generalization ability of these methods is limited. In this paper, we propose a sparse boosting representation classification (SBRC) for robust face recognition. To improve the effectiveness of representation coding, an error detection machine (EDM) with multiple error detectors (ED) in SBRC, is proposed to detect and remove destroyed features (i.e. pixels) on a testing image. SBRC has three advantages: First, it has good generalization ability, since the EDM can self-adjust the number of ED according to different variations; Second, EDM would boost the sparsity of coding vector; Third, its implementation is simple and efficient as the EDM is based on l 2 − n o r m . In addition, five popular face image databases including AR database, Extended Yale B database, ORL database, FERET database and LFW database were applied to validate the performance of SBRC. The superiority of SBRC is confirmed by comparing it with the state-of-the-art face recognition methods. [ABSTRACT FROM AUTHOR]

In sparse representation (SR), a test image is encoded by a sparse linear combination of training samples. The L1‐regulariser used in SR is beneficial to produce a good reconstruction of the test face image with sparse error, but it is incapable to guarantee the robustness against local structural noise. To enhance noise tolerance of SR‐based classifier, an improved L1‐regulariser based on trimmed sparse coding (TSC) by using an extra penalty on correlation among all coding coefficients is proposed. Different from traditional single‐coding scheme in SR, multiple coding coefficients are used to represent patches of a test image by its corresponding training patches. The consistency penalty imposed into the new SR model improves the confidence for accuracy classification. Experimental results show the superiority of TSC on two benchmark databases, and it outperforms other state‐of‐the‐art methods. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, we present a comparative assessment of two ICA using reference signal methods. Independent Component Analysis (ICA) using reference signal is a useful tool for extracting a desired independent component (IC). Reference signal is served as a priori information to conduct the one-unit ICA to converge to the local extreme point related to a desired IC. There are two methods can perform ICA using reference signal, namely ICA with reference (ICA-R) and fast ICA with reference signal (FICAR). This paper intends to fill in a gap in the previous studies to give comparisons of those methods systematically. Moreover, we also provide useful improvements to the two methods. Firstly, we propose a improved algorithm to fix the flaw in the previous FICAR. Secondly, for ICA-R, we propose a criterion to select appropriate distance measurement. [Copyright &y& Elsevier]

This paper presents a stable and fast algorithm for independent component analysis with reference (ICA-R). This is a technique for incorporating available reference signals into the ICA contrast function so as to form an augmented Lagrangian function under the framework of constrained ICA (cICA). The previous ICA-R algorithm was constructed by solving the optimization problem via a Newton-like learning style. Unfortunately, the slow convergence and potential misconvergence limit the capability of ICA-R. This paper first investigates and probes the flaws of the previous algorithm and then introduces a new stable algorithm with a faster convergence speed. There are two other highlights in this paper: first, new approaches, including the reference deflation technique and a direct way of obtaining references, are introduced to facilitate the application of ICA-R; second, a new method is proposed that the new ICA-R is used to recover the complete underlying sources with new advantages compared with other classical ICA methods. Finally, the experiments on both synthetic and real-world data verify the better performance of the new algorithm over both previous ICA-R and other well-known methods. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, we propose a novel method to recognize the partially occluded face images based on sparse representation. Generally, occlusions, such as glasses and scarf, fall on some random patch of image's pixels of test images, but which is known to be connected. In our method, all images are divided into several blocks and then an indicator based on linear regression technique is presented to determine whether a block is occluded. We utilize those uncontaminated blocks as the new feature of an image. Finally, the sparse representation-based classification (SRC) method serves as the classifier to recognize unknown faces. In the original work of SRC, the extended SRC (eSRC) scheme is presented to deal with occlusions, which has very heavy computational cost. The experimental results show that our method can achieve good recognition accuracy and has much lower computational cost than eSRC. [Copyright &y& Elsevier]

Abstract: In this paper, we present a collaborative representation-based classification on selected training samples (CRC_STS) for face image recognition. The CRC_STS uses a two stage scheme: The first stage is to select some most significant training samples from the original training set by using a multiple round of refining process. The second stage is to use collaborative representation classifier to perform classification on the selected training samples. Our method can be regarded as a sparse representation approach but without imposing l 1-norm constraint on representation coefficients. The experimental results on three well known face databases show that our method works very well. [Copyright &y& Elsevier]

Abstract: The nearest subspace (NS) classification is an efficient method to solve face recognition problem by using the linear regression technique. This method is based on the assumption that face images from a specific subject class tend to span a unique subspace, i.e. a class-specific subspace. Then, a test image has the shortest distance from its own class-specific subspace. In this paper, we present a novel idea for face recognition. This idea considers that a test face image should be far from the farthest subspace (FS) spanned by all training images except the images from the class of this test image. Based on this idea, we propose the FS classifier for face recognition. In our opinion, NS and FS classifiers take advantages of different characteristics of the class-specific subspace. NS classifier exploits the relationship between a test image and a single class while FS classifier measures relationship between this test image and the rest classes. Consequently, we propose the nearest-farthest subspace (NFS) classifier which exploits the both relationships to classify a test image. The comparisons with NS classifier and other state-of-the-art methods on four famous public face databases demonstrate the good performance of FS and NFS. [Copyright &y& Elsevier]

The sparse representation-based classification (SRC) has been proven to be a robust face recognition method. However, its computational complexity is very high due to solving a complex -minimization problem. To improve the calculation efficiency, we propose a novel face recognition method, called sparse representation-based classification on k-nearest subspace (SRC-KNS). Our method first exploits the distance between the test image and the subspace of each individual class to determine the nearest subspaces and then performs SRC on the selected classes. Actually, SRC-KNS is able to reduce the scale of the sparse representation problem greatly and the computation to determine the nearest subspaces is quite simple. Therefore, SRC-KNS has a much lower computational complexity than the original SRC. In order to well recognize the occluded face images, we propose the modular SRC-KNS. For this modular method, face images are partitioned into a number of blocks first and then we propose an indicator to remove the contaminated blocks and choose the nearest subspaces. Finally, SRC is used to classify the occluded test sample in the new feature space. Compared to the approach used in the original SRC work, our modular SRC-KNS can greatly reduce the computational load. A number of face recognition experiments show that our methods have five times speed-up at least compared to the original SRC, while achieving comparable or even better recognition rates. [ABSTRACT FROM AUTHOR]

Face recognition in harsh environments is an active research topic. As one of the most important challenges, face recognition across pose has received extensive attention. LBP feature has been used widely in face recognition because of its robustness to slight illumination and pose variations. However, due to the way of pattern feature calculation, its effectiveness is limited by the big rotations. In this paper, a new LBP-like feature extraction is proposed which modifies the code rule by Huffman. Besides, a Divide-and-Rule strategy is applied to both face representation and classification, which aims to improve recognition performance across pose. Extensive experiments on CMU PIE database, FERET database and LFW database are conducted to verify the efficacy of the proposed method. The experimental results show that our method significantly outperforms other approaches. [ABSTRACT FROM AUTHOR]

Conventional representation-based methods only consider the whole training set for representation and recognition. However, some of the training samples make little contribution to the representation of the test sample. In this paper, we propose to construct an optimal representation set for recognition. First, the nearest neighbour principle is used to initialize the optimal representation set. Then, it will be updated through adding a training sample which can work with the current representation set to represent the test sample with minimum error. With this scheme, we accelerate the computation procedure by the partitioned matrix technique. To fully validate the effectiveness of the proposed method, experiments have been conducted on public palmprint and face databases by comparing the proposed method with the state-of-the-art methods. The proposed competitive sample selection method shows promising results. [ABSTRACT FROM PUBLISHER]

In terms of making genes expression data more interpretable and comprehensible, there exists a significant superiority on sparse methods. Many sparse methods, such as penalized matrix decomposition (PMD) and sparse principal component analysis (SPCA), have been applied to extract plants core genes. Supervised algorithms, especially the support vector machine-recursive feature elimination (SVM-RFE) method, always have good performance in gene selection. In this paper, we draw into class information via the total scatter matrix and put forward a class-information-based penalized matrix decomposition (CIPMD) method to improve the gene identification performance of PMD-based method. Firstly, the total scatter matrix is obtained based on different samples of the gene expression data. Secondly, a new data matrix is constructed by decomposing the total scatter matrix. Thirdly, the new data matrix is decomposed by PMD to obtain the sparse eigensamples. Finally, the core genes are identified according to the nonzero entries in eigensamples. The results on simulation data show that CIPMD method can reach higher identification accuracies than the conventional gene identification methods. Moreover, the results on real gene expression data demonstrate that CIPMD method can identify more core genes closely related to the abiotic stresses than the other methods. [ABSTRACT FROM AUTHOR]