Search

The purpose of object counting is to estimate the number of specific kinds of objects in a given image. In remote-sensing imagery, challenges arise in object counting due to issues like scale variations and complex backgrounds. Existing density map-based object counting methods have achieved satisfactory performance in some general scenarios (i.e., crowd counting and vehicle counting) and have become the mainstream methods. These density map-based counting methods use a fixed Gaussian kernel in the density map generation stage, thus they are not well adapted to the challenges such as scale variations present in remote-sensing scenes. In this letter, we propose to use the strategy of neural architecture search (NAS-Kernel) to select appropriate Gaussian kernels corresponding to objects of different scales in the Gaussian density map generation stage. NAS-Kernel is a plug-and-play algorithm that can be used in other density map-based counting methods. In addition, a contextual path aggregation (CPA) feature fusion strategy is proposed to fuse multiscale feature information. The ablation experiments verify that the proposed method can significantly improve the performance of the baseline. Experimental results on the four subdatasets of RSOC show that the proposed method achieves state-of-the-art performance. On the Building subdataset, the proposed method achieves 18% and 12% lower mean absolute error (MAE) and root mean square error (RMSE) than the existing methods. [ABSTRACT FROM AUTHOR]

Dimension reduction is a critical technology for high-dimensional data processing, where Linear Discriminant Analysis (LDA) and its variants are effective supervised methods. However, LDA prefers to feature with smaller variance, which causes feature with weak discriminative ability retained. In this paper, we propose a novel Ratio Sum for Linear Discriminant Analysis (RSLDA), which aims at maximizing discriminative ability of each feature in subspace. To be specific, it maximizes the sum of ratio of the between-class distance to the within-class distance in each dimension of subspace. Since the original RSLDA problem is difficult to obtain the closed solution, an equivalent problem is developed which can be solved by an alternative optimization algorithm. For solving the equivalent problem, it is transformed into two sub-problems, one of which can be solved directly, the other is changed into a convex optimization problem, where singular value decomposition is employed instead of matrix inversion. Consequently, performance of algorithm cannot be affected by the non-singularity of covariance matrix. Furthermore, Kernel RSLDA (KRSLDA) is presented to improve the robustness of RSLDA. Additionally, time complexity of RSLDA and KRSLDA are analyzed. Extensive experiments show that RSLDA and KRSLDA outperforms other comparison methods on toy datasets and multiple public datasets. [ABSTRACT FROM AUTHOR]

Binarized neural networks (BNNs) have drawn significant attention in recent years, owing to great potential in reducing computation and storage consumption. While it is attractive, traditional BNNs usually suffer from slow convergence speed and dramatical accuracy-degradation on large-scale classification datasets. To minimize the gap between BNNs and deep neural networks (DNNs), we propose a new framework of designing BNNs, dubbed Hyper-BinaryNet, from the aspect of enhanced information-flow. Our contributions are threefold: 1) Considering the capacity-limitation in the backward pass, we propose an 1-bit convolution module named HyperConv. By exploiting the capacity of auxiliary neural networks, BNNs gain better performance on large-scale image classification task. 2) Considering the slow convergence speed in BNNs, we rethink the gradient accumulation mechanism and propose a hyper accumulation technique. By accumulating gradients in multiple variables rather than one as before, the gradient paths for each weight increase, which escapes BNNs from the gradient bottleneck problem during training. 3) Considering the ill-posed optimization problem, a novel gradient estimation warmup strategy, dubbed STE-Warmup, is developed. This strategy prevents BNNs from the unstable optimization process by progressively transferring neural networks from 32-bit to 1-bit. We conduct evaluations with variant architectures on three public datasets: CIFAR-10/100 and ImageNet. Compared with state-of-the-art BNNs, Hyper-BinaryNet shows faster convergence speed and outperforms existing BNNs by a large margin. [ABSTRACT FROM AUTHOR]

As a multivariate data analysis tool, canonical correlation analysis (CCA) has been widely used in computer vision and pattern recognition. However, CCA uses Euclidean distance as a metric, which is sensitive to noise or outliers in the data. Furthermore, CCA demands that the two training sets must have the same number of training samples, which limits the performance of CCA-based methods. To overcome these limitations of CCA, two novel canonical correlation learning methods based on low-rank learning are proposed in this paper for image representation, named robust canonical correlation analysis (robust-CCA) and low-rank representation canonical correlation analysis (LRR-CCA). By introducing two regular matrices, the training sample numbers of the two training datasets can be set as any values without any limitation in the two proposed methods. Specifically, robust-CCA uses low-rank learning to remove the noise in the data and extracts the maximization correlation features from the two learned clean data matrices. The nuclear norm and $L_{1}$ -norm are used as constraints for the learned clean matrices and noise matrices, respectively. LRR-CCA introduces low-rank representation into CCA to ensure that the correlative features can be obtained in low-rank representation. To verify the performance of the proposed methods, five publicly image databases are used to conduct extensive experiments. The experimental results demonstrate the proposed methods outperform state-of-the-art CCA-based and low-rank learning methods. [ABSTRACT FROM AUTHOR]

The multiple kernel $k$ -means (MKKM) and its variants utilize complementary information from different sources, achieving better performance than kernel $k$ -means (KKM). However, the optimization procedures of most previous works comprise two stages, learning the continuous relaxation matrix and obtaining the discrete one by extra discretization procedures. Such a two-stage strategy gives rise to a mismatched problem and severe information loss. Even worse, most existing MKKM methods overlook the correlation among prespecified kernels, which leads to the fusion of mutually redundant kernels and bad effects on the diversity of information sources, finally resulting in unsatisfying results. To address these issues, we elaborate a novel Discrete and Parameter-free Multiple Kernel $k$ -means (DPMKKM) model solved by an alternative optimization method, which can directly obtain the cluster assignment results without subsequent discretization procedure. Moreover, DPMKKM can measure the correlation among kernels by implicitly introducing a regularization term, which is able to enhance kernel fusion by reducing redundancy and improving diversity. Noteworthily, the time complexity of optimization algorithm is successfully reduced, through masterly utilizing of coordinate descent technique, which contributes to higher algorithm efficiency and broader applications. What’s more, our proposed model is parameter-free avoiding intractable hyperparameter tuning, which makes it feasible in practical applications. Lastly, extensive experiments conducted on a number of real-world datasets illustrated the effectiveness and superiority of the proposed DPMKKM model. [ABSTRACT FROM AUTHOR]

Traditional graph-based Semi-Supervised Learning (SSL) methods usually contain two separate steps. First, constructing an affinity matrix. Second, inferring the unknown labels. While such a two-step method has been successful, it cannot take full advantage of the correlation between affinity matrix and label information. In order to address the above problem, we propose a novel graph-based SSL method. It can learn the affinity matrix and infer the unknown labels simultaneously. Moreover, feature selection with auto-weighting is introduced to extract the effective and robust features. Further, the proposed method learns the data similarity matrix by assigning the adaptive neighbors for each data point based on the local distance. We solve the unified problem via an alternative minimization algorithm. Extensive experimental results on synthetic data and benchmark data show that the proposed method consistently outperforms the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

This paper addresses two fundamental problems: 1) learning discriminative model parameters and 2) avoiding over-fitting, which often occurs in regression-based classification tasks. We formulate these two problems in terms of relaxing both the strict binary label matrix and graph regularization term into more flexible forms so that the margins between different classes are enlarged as much as possible and the problem of over-fitting is avoided to some extent. This task is accomplished by the proposed double relaxed regression (DRR) method. The convex problem of DRR is solved efficiently with an iterative procedure. Extensive experiments on synthetic and real world image data sets demonstrate the effectiveness of the proposed method in terms of both classification accuracy and running time. [ABSTRACT FROM AUTHOR]

Spectral clustering is an important clustering method widely used for pattern recognition and image segmentation. Classical spectral clustering algorithms consist of two separate stages: 1) solving a relaxed continuous optimization problem to obtain a real matrix followed by 2) applying ${K}$ -means or spectral rotation to round the real matrix (i.e., continuous clustering result) into a binary matrix called the cluster indicator matrix. Such a separate scheme is not guaranteed to achieve jointly optimal result because of the loss of useful information. To obtain a better clustering result, in this paper, we propose a joint model to simultaneously compute the optimal real matrix and binary matrix. The existing joint model adopts an orthonormal real matrix to approximate the orthogonal but nonorthonormal cluster indicator matrix. It is noted that only in a very special case (i.e., all clusters have the same number of samples), the cluster indicator matrix is an orthonormal matrix multiplied by a real number. The error of approximating a nonorthonormal matrix is inevitably large. To overcome the drawback, we propose replacing the nonorthonormal cluster indicator matrix with a scaled cluster indicator matrix which is an orthonormal matrix. Our method is capable of obtaining better performance because it is easy to minimize the difference between two orthonormal matrices. Experimental results on benchmark datasets demonstrate the effectiveness of the proposed method (called JSESR). [ABSTRACT FROM AUTHOR]

Preserving projection-based methods are good for finding the manifold structure embedded in data. As they use the Euclidean distance as a metric, which is sensitive to noise and outliers in data, nuclear norm-based 2D locality preserving projection (NN-2DLPP) is thus proposed to improve the robustness of 2DLPP. However, NN-2DLPP does not consider the discriminant ability of data. In order to improve the discriminant ability of preserving projection methods, in this paper, we use preserving projection learning with structurally incoherence of data and propose structurally incoherent low-rank 2DLPP (SILR-2DLPP) for image classification. This approach provides a discriminative representation of preserving projection learning by recovering the distinct different classes of the data. SILR-2DLPP searches the optimal subspace and low-rank representation simultaneously. We further extend SILR-2DLPP to a kernel case and propose kernel SILR-2DLPP (KSILR-2DLPP) to obtain a nonlinear representation. The theoretical analysis including the convergence and computational complexity of SILR-2DLPP are presented. To verify the performance of SILR-2DLPP and KSILR-2DLPP, six well-known image databases were used in the experiments. The experimental results show that the proposed methods are superior to the previous preserving projection methods for image classification. [ABSTRACT FROM AUTHOR]

This paper presents a hybrid level set method for object segmentation. The method deconstructs segmentation task into two procedures, i.e., shape transformation and curve evolution, which are alternately optimized until convergence. In this framework, only one shape prior encoded by shape context is utilized to estimate a transformation allowing the curve to have the same semantic expression as shape prior, and curve evolution is driven by an energy functional with topology-preserving and kernelized terms. In such a way, the proposed method is featured by the following advantages: 1) hybrid paradigm makes the level set framework possess the ability of incorporating other shape-related techniques about shape descriptor and distance; 2) shape context endows one single prior with semanticity, and hence leads to the competitive performance compared to the ones with multiple shape priors; and 3) additionally, combining topology-preserving and kernelization mechanisms together contributes to realizing a more reasonable segmentation on textured and noisy images. As far as we know, we propose a hybrid level set framework and utilize shape context to guide curve evolution for the first time. Our method is evaluated with synthetic, healthcare, and natural images, as a result, it shows competitive and even better performance compared to the counterparts. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) classification, which aims to assign an accurate label for hyperspectral pixels, has drawn great interest in recent years. Although low-rank representation (LRR) has been used to classify HSI, its ability to segment each class from the whole HSI data has not been exploited fully yet. LRR has a good capacity to capture the underlying low-dimensional subspaces embedded in original data. However, there are still two drawbacks for LRR. First, the LRR does not consider the local geometric structure within data, which makes the local correlation among neighboring data easily ignored. Second, the representation obtained by solving LRR is not discriminative enough to separate different data. In this paper, a novel locality- and structure-regularized LRR (LSLRR) model is proposed for HSI classification. To overcome the above-mentioned limitations, we present locality constraint criterion and structure preserving strategy to improve the classical LRR. Specifically, we introduce a new distance metric, which combines both spatial and spectral features, to explore the local similarity of pixels. Thus, the global and local structures of HSI data can be exploited sufficiently. In addition, we propose a structural constraint to make the representation have a near-block-diagonal structure. This helps to determine the final classification labels directly. Extensive experiments have been conducted on three popular HSI data sets. And the experimental results demonstrate that the proposed LSLRR outperforms other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Blur detection in a single image is challenging especially when the blur is spatially-varying. Developing discriminative blur features is an open problem. In this paper, we propose a new kernel-specific feature vector consisting of the information of a blur kernel and the information of an image patch. Specifically, the kernel specific-feature is composed of the multiplication of the variance of filtered kernel and the variance of filtered patch gradients. The feature origins from a blur-classification theorem and its discrimination can also be intuitively explained. To make the kernel-specific features useful for real applications, we build a pool of kernels consisting of motion-blur kernels, defocus-blur (out-of-focus) kernels, and their combinations. By extracting such features followed by the classifiers, the proposed algorithm outperforms the state-of-the-art blur detection method. Experimental results on public databases demonstrate the effectiveness of the proposed method. [ABSTRACT FROM PUBLISHER]

Motion is an important clue for industrial inspection, video surveillance, and service machines to localize and recognize products and objects. Because blur co-occurs with motion, it is desirable for developing efficient and robust motion blur detection algorithm. However, existing algorithms are inefficient for detecting spatially varying motion blur. To deal with the problem, this paper presents a theorem, according to which, motion blur can be efficiently detected and segmented. According to the theorem, the proposed algorithm requires a simple filtering operation and variance computation. Classification as either blurred or unblurred pixel can be done by substituting the variance into the proposed simple formula and checking the sign of the resulting value. Moreover, a geometric interpretation and two extensions of the algorithm are given. Importantly, based on the geometric interpretation of the indicator function, we develop a one-class classifier, which is more effective than the indicator function and has comparable computational cost of the indicator function. Experimental results on detecting motion-blurred cars, motorcycles, bicycles, bags, and persons demonstrate that the proposed algorithm is very efficient without loss of effectiveness. [ABSTRACT FROM PUBLISHER]

Fast and accurately categorizing the millions of aerial images on Google Maps is a useful technique in pattern recognition. Existing methods cannot handle this task successfully due to two reasons: 1) the aerial images’ topologies are the key feature to distinguish their categories, but they cannot be effectively encoded by a conventional visual codebook and 2) it is challenging to build a realtime image categorization system, as some geo-aware Apps update over 20 aerial images per second. To solve these problems, we propose an efficient aerial image categorization algorithm. It focuses on learning a discriminative topological codebook of aerial images under a multitask learning framework. The pipeline can be summarized as follows. We first construct a region adjacency graph (RAG) that describes the topology of each aerial image. Naturally, aerial image categorization can be formulated as RAG-to-RAG matching. According to graph theory, RAG-to-RAG matching is conducted by enumeratively comparing all their respective graphlets (i.e., small subgraphs). To alleviate the high time consumption, we propose to learn a codebook containing topologies jointly discriminative to multiple categories. The learned topological codebook guides the extraction of the discriminative graphlets. Finally, these graphlets are integrated into an AdaBoost model for predicting aerial image categories. Experimental results show that our approach is competitive to several existing recognition models. Furthermore, over 24 aerial images are processed per second, demonstrating that our approach is ready for real-world applications. [ABSTRACT FROM PUBLISHER]

Fine-grained image categorization is a challenging task aiming at distinguishing objects belonging to the same basic-level category, e.g., leaf or mushroom. It is a useful technique that can be applied for species recognition, face verification, and so on. Most of the existing methods either have difficulties to detect discriminative object components automatically, or suffer from the limited amount of training data in each sub-category. To solve these problems, this paper proposes a new fine-grained image categorization model. The key is a dense graph mining algorithm that hierarchically localizes discriminative object parts in each image. More specifically, to mimic the human hierarchical perception mechanism, a superpixel pyramid is generated for each image. Thereby, graphlets from each layer are constructed to seamlessly capture object components. Intuitively, graphlets representative to each super-/sub-category is densely distributed in their feature space. Thus, a dense graph mining algorithm is developed to discover graphlets representative to each super-/sub-category. Finally, the discovered graphlets from pairwise images are integrated into an image kernel for fine-grained recognition. Theoretically, the learned kernel can generalize several state-of-the-art image kernels. Experiments on nine image sets demonstrate the advantage of our method. Moreover, the discovered graphlets from each sub-category accurately capture those tiny discriminative object components, e.g., bird claws, heads, and bodies. [ABSTRACT FROM PUBLISHER]

We present a framework for image classification that extends beyond the window sampling of fixed spatial pyramids and is supported by a new learning algorithm. Based on the observation that fixed spatial pyramids sample a rather limited subset of the possible image windows, we propose a method that accounts for a comprehensive set of windows densely sampled over location, size, and aspect ratio. A concise high-level image feature is derived to effectively deal with this large set of windows, and this higher level of abstraction offers both efficient handling of the dense samples and reduced sensitivity to misalignment. In addition to dense window sampling, we introduce generalized adaptive \ell p -norm multiple kernel learning (GA-MKL) to learn a robust classifier based on multiple base kernels constructed from the new image features and multiple sets of prelearned classifiers from other classes. With GA-MKL, multiple levels of image features are effectively fused, and information is shared among different classifiers. Extensive evaluation on benchmark datasets for object recognition (Caltech256 and Caltech101) and scene recognition (15Scenes) demonstrate that the proposed method outperforms the state-of-the-art under a broad range of settings. [ABSTRACT FROM PUBLISHER]

We propose a new graph based hashing method called spectral embedded hashing (SEH) for large-scale image retrieval. We first introduce a new regularizer into the objective function of the recent work spectral hashing to control the mismatch between the resultant hamming embedding and the low-dimensional data representation, which is obtained by using a linear regression function. This linear regression function can be employed to effectively handle the out-of-sample data, and the introduction of the new regularizer makes SEH better cope with the data sampled from a nonlinear manifold. Considering that SEH cannot efficiently cope with the high dimensional data, we further extend SEH to kernel SEH (KSEH) to improve the efficiency and effectiveness, in which a nonlinear regression function can also be employed to obtain the low dimensional data representation. We also develop a new method to efficiently solve the approximate solution for the eigenvalue decomposition problem in SEH and KSEH. Moreover, we show that some existing hashing methods are special cases of our KSEH. Our comprehensive experiments on CIFAR, Tiny-580K, NUS-WIDE, and Caltech-256 datasets clearly demonstrate the effectiveness of our methods. [ABSTRACT FROM PUBLISHER]

Image super-resolution (SR) reconstruction is essentially an ill-posed problem, so it is important to design an effective prior. For this purpose, we propose a novel image SR method by learning both non-local and local regularization priors from a given low-resolution image. The non-local prior takes advantage of the redundancy of similar patches in natural images, while the local prior assumes that a target pixel can be estimated by a weighted average of its neighbors. Based on the above considerations, we utilize the non-local means filter to learn a non-local prior and the steering kernel regression to learn a local prior. By assembling the two complementary regularization terms, we propose a maximum a posteriori probability framework for SR recovery. Thorough experimental results suggest that the proposed SR method can reconstruct higher quality results both quantitatively and perceptually. [ABSTRACT FROM AUTHOR]

The Gaussian process latent variable model (GP-LVM) has been identified to be an effective probabilistic approach for dimensionality reduction because it can obtain a low-dimensional manifold of a data set in an unsupervised fashion. Consequently, the GP-LVM is insufficient for supervised learning tasks (e.g., classification and regression) because it ignores the class label information for dimensionality reduction. In this paper, a supervised GP-LVM is developed for supervised learning tasks, and the maximum a posteriori algorithm is introduced to estimate positions of all samples in the latent variable space. We present experimental evidences suggesting that the supervised GP-LVM is able to use the class label information effectively, and thus, it outperforms the GP-LVM and the discriminative extension of the GP-LVM consistently. The comparison with some supervised classification methods, such as Gaussian process classification and support vector machines, is also given to illustrate the advantage of the proposed method. [ABSTRACT FROM AUTHOR]

In the field of machine learning, it is a key issue to learn and represent similarity. This paper focuses on the problem of learning similarity with a multikernel method. Motivated by geometric intuition and computability, similarity between patterns is proposed to be measured by their included angle in a kernel-induced Hilbert space. Having noticed that the cosine of such an included angle can be represented by a normalized kernel, it can be said that the task of learning similarity is equivalent to learning an appropriate normalized kernel. In addition, an error bound is also established for learning similarity with the multikernel method. Based on this bound, a boosting-style algorithm is developed. The preliminary experiments validate the effectiveness of the algorithm for learning similarity. [ABSTRACT FROM AUTHOR]