Search

When the training data and the test data do not obey the same distribution, the performances of many object detection methods always decrease greatly. Naturally, domain adaptation methods at feature level are proposed. The basic idea is to adapt the feature extraction network such that the feature distributions of the source and target domains match. We propose a new method that is built directly on the Faster R-CNN model, which not only aligns the source and target data features, but also forces their generated features closer together to further align the source and target domains. Moreover, compared with previous approaches, we construct a more powerful discriminator and a simple generator to solve the domain adaptation problem. The model works like a pair of scissors, so we call it Scissors Networks (SN). We conduct extensive experiments on popular datasets, including Cityscapes, Foggy Cityscapes, SIM10k and KITTI. The experimental results demonstrate that our algorithm is superior to the state-of-the-art deep learning based domain adaptation approaches. [ABSTRACT FROM AUTHOR]

Feature selection is a widely used dimension reduction technique to select feature subsets because of its interpretability. Many methods have been proposed and achieved good results, in which the relationships between adjacent data points are mainly concerned. But the possible associations between data pairs that are not adjacent are always neglected. Different from previous methods, we propose a novel and very simple approach for unsupervised feature selection, named MMFS (Multi-step Markov Probability Relationship for Feature Selection). The idea is using multi-step Markov transition probability to describe the relation between any data pair. Two ways from the positive and negative viewpoints are employed respectively to keep the data structure after feature selection. From the positive viewpoint, the maximum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. Then, the features which can keep the compact data structure are selected. From the viewpoint of negative, the minimum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. On the contrary, the features that least maintain the loose data structure are selected. The two ways can also be combined. Thus three algorithms are proposed. Our main contributions are a novel feature section approach which uses multi-step transition probability to characterize the data structure, and three algorithms proposed from the positive and negative aspects for keeping data structure and select the features to preserve such structure. The performance of our approach is compared with the state-of-the-art methods on eight real-world data sets, and the experimental results show that the proposed MMFS is effective in unsupervised feature selection. [ABSTRACT FROM AUTHOR]

We present a practical and inexpensive method to reconstruct 3D scenes that include transparent and mirror objects. Our work is motivated by the need for automatically generating 3D models of interior scenes, which commonly include glass. These large structures are often invisible to cameras or even to our human visual system. Existing 3D reconstruction methods for transparent objects are usually not applicable in such a room-sized reconstruction setting. Our simple hardware setup augments a regular depth camera (e.g., the Microsoft Kinect camera) with a single ultrasonic sensor, which is able to measure the distance to any object, including transparent surfaces. The key technical challenge is the sparse sampling rate from the acoustic sensor, which only takes one point measurement per frame. To address this challenge, we take advantage of the fact that the large scale glass structures in indoor environments are usually either piece-wise planar or a simple parametric surface. Based on these assumptions, we have developed a novel sensor fusion algorithm that first segments the (hybrid) depth map into different categories such as opaque/transparent/infinity (e.g., too far to measure) and then updates the depth map based on the segmentation outcome. We validated our algorithms with a number of challenging cases, including multiple panes of glass, mirrors, and even a curved glass cabinet. [ABSTRACT FROM AUTHOR]

In this paper we present a novel real-time algorithm for simultaneous pose and shape estimation for articulated objects, such as human beings and animals. The key of our pose estimation component is to embed the articulated deformation model with exponential-maps-based parametrization into a Gaussian Mixture Model. Benefiting from this probabilistic measurement model, our algorithm requires no explicit point correspondences as opposed to most existing methods. Consequently, our approach is less sensitive to local minimum and handles fast and complex motions well. Moreover, our novel shape adaptation algorithm based on the same probabilistic model automatically captures the shape of the subjects during the dynamic pose estimation process. The personalized shape model in turn improves the tracking accuracy. Furthermore, we propose novel approaches to use either a mesh model or a sphere-set model as the template for both pose and shape estimation under this unified framework. Extensive evaluations on publicly available data sets demonstrate that our method outperforms most state-of-the-art pose estimation algorithms with large margin, especially in the case of challenging motions. Furthermore, our shape estimation method achieves comparable accuracy with state of the arts, yet requires neither statistical shape model nor extra calibration procedure. Our algorithm is not only accurate but also fast, we have implemented the entire processing pipeline on GPU. It can achieve up to 60 frames per second on a middle-range graphics card. [ABSTRACT FROM PUBLISHER]

In this paper, we introduce the notion of sufficient set and necessary set for distributed processing of probabilistic top-k queries in cluster-based wireless sensor networks. These two concepts have very nice properties that can facilitate localized data pruning in clusters. Accordingly, we develop a suite of algorithms, namely, sufficient set-based (SSB), necessary set-based (NSB), and boundary-based (BB), for intercluster query processing with bounded rounds of communications. Moreover, in responding to dynamic changes of data distribution in the network, we develop an adaptive algorithm that dynamically switches among the three proposed algorithms to minimize the transmission cost. We show the applicability of sufficient set and necessary set to wireless sensor networks with both two-tier hierarchical and tree-structured network topologies. Experimental results show that the proposed algorithms reduce data transmissions significantly and incur only small constant rounds of data communications. The experimental results also demonstrate the superiority of the adaptive algorithm, which achieves a near-optimal performance under various conditions. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, the multistability of a class of Amari’s -divergence based nonnegative matrix factorization learning algorithms is analyzed. The analysis results show that invariant sets for the update algorithms can be constructed. In these invariant sets, the non-convergence of the discussed algorithms can be guaranteed. Based on Lyapunov’s stability theorem, the local convergence of this class of learning algorithms is proved in the domain of their update rules. In the simulation, the analysis results are applied to image representation. Experiment results demonstrate that selecting suitable initial data for different applications of these nonnegative matrix factorization algorithms is very important. [Copyright &y& Elsevier]

Abstract: Detecting outlier efficiently is an active research issue in data mining, which has important applications in the field of fraud detection, network intrusion detection, monitoring criminal activities in electronic commerce, etc. Because of the sparsity of high dimensional data, it is reasonable and meaningful to detect the outliers in suitable projected subspaces. We call such subspace and outliers in the subspace as anomaly subspace and projected outlier respectively. Many efficient algorithms have already been proposed for outlier detection based on different approaches, but there are few literatures on projected outlier detection for high dimensional data sets with mixed continuous and categorical attributes. In this paper, a novel projected outlier detection algorithm is proposed to detect projected outliers in high-dimensional mixed attribute data set. Our main contributions are: (1) combined with information entropy, a novel measure of anomaly subspace is proposed. In this anomaly subspace, meaningful outliers could be detected and explained. Unlike the previous projected outlier detection methods, the dimension of anomaly subspace is not decided beforehand; (2) theoretical analysis about this measure is presented; (3) bottom-up method is proposed to find the interesting anomaly subspaces; (4) the outlying degree of projected outlier is defined, which has good explanations; (5) the data set with mixed data type is handled; (6) experiments on synthetic and real data sets to evaluate the effectiveness of our approach are performed. [Copyright &y& Elsevier]

When the independent sources are known to be nonnegative and well-grounded, which means that they have a nonzero pdf in the region of zero, Oja and Plumbley have proposed a "Nonnegative principal component analysis (PCA)" algorithm to separate these positive sources. Generally, it is very difficult to prove the convergence of a discrete-time independent component analysis (ICA) learning algorithm. However, by using the skew-symmetry property of this discrete-time "Nonnegative PCA" algorithm, if the learning rate satisfies suitable condition, the global convergence of this discrete-time algorithm can be proven. Simulation results are employed to further illustrate the advantages of this theory. [ABSTRACT FROM AUTHOR]

Abstract: On minor component analysis (MCA) neural networks, a new algorithm is proposed. It is a self-stabilizing MCA algorithm, which means that the sign of the temporal change of the weight vector length is independent of the presented input vector. Algorithms without this property may suffer fluctuations and divergence. With suitable conditions on the initial weight vector and learning rate, a rigorous global convergence proof is given. The techniques used in the proof will be useful in many research issues such as independent component analysis, principle component analysis, etc. [Copyright &y& Elsevier]

Stochastic gradient Markov chain Monte Carlo algorithms have received much attention in Bayesian computing for big data problems, but they are only applicable to a small class of problems for which the parameter space has a fixed dimension and the log-posterior density is differentiable with respect to the parameters. This paper proposes an extended stochastic gradient Markov chain Monte Carlo algorithm which, by introducing appropriate latent variables, can be applied to more general large-scale Bayesian computing problems, such as those involving dimension jumping and missing data. Numerical studies show that the proposed algorithm is highly scalable and much more efficient than traditional Markov chain Monte Carlo algorithms. [ABSTRACT FROM AUTHOR]

The precise positioning of a lander is a key issue during in situ planetary exploration. This paper proposes an optimal weighting algorithm based on a Helmert-variance component estimation applied to VLBI phase delay data. We validated our algorithm on Chang'e 3 lander VLBI phase delay from the Chinese VLBI Network. We consider that the lander positioning accuracy achieved with our new algorithm is at least 53 % better, compared to previous estimations based on the same observations. The position of the lander based on continuous 4-day observations was estimated to be 44.1215 ° N (± 21.9 m), −19.5135 ° E (± 4.2 m) and −2632.7 m (± 18.9 m) with respect to a moon reference system. This result agrees well with the position derived from images taken by the Narrow Angle Camera on the Lunar Reconnaissance Orbiter. [ABSTRACT FROM AUTHOR]

Image reconstruction plays a key role in the application of electrical capacitance tomography (ECT). Although many different algorithms have been developed in the past, it is often difficult to obtain satisfactory images in all imaging regions by the use of a single algorithm due to the soft-field nature of ECT. This motivated us to develop a novel ECT image reconstruction strategy, in which a high-quality image can be obtained by combining the images reconstructed by two different algorithms via a graph cut method. By doing so, it is possible to retain the advantage of each algorithm for specified imaging regions and improve the quality of the whole image. This strategy was verified both numerically and experimentally for two widely used ECT reconstruction algorithms, i.e., linear backprojection and Tikhonov regularization. The results for stationary objects as well as gas–solid fluidized beds demonstrated that this graph-cut-based combination strategy presents a promising approach for ECT image reconstruction. [ABSTRACT FROM AUTHOR]

Moving object detection is about foreground and background separation based on motion detection. Detecting moving objects from similarly colored background (known as camouflage problem) has been a long-standing open question in this field. Discriminative modeling (DM), which focuses on enhancing the performance to distinguish foreground from background with discriminative features and well-designed classifiers, has been widely used for moving object detection. However, DM may tend to fail when encountering the camouflage problem, as the class separability in camouflaged areas is generally poor. In this paper, we propose a new strategy, camouflage modeling (CM), to identify camouflaged foreground pixels. In view of the fact that camouflage involves both foreground and background, we need to model both the background and the foreground, and compare them in a well-designed way in camouflage detection. Specifically, we develop a global model for the background, and an integration of global and local models for the foreground, respectively. Based on both background and foreground models, we introduce a factor to measure the degree of camouflage, and further identify truly camouflaged areas. In view of the fact that a moving object is usually composed of both camouflaged and noncamouflaged areas, CM and DM are fused in a Bayesian framework to perform complete object detection. Experiments are conducted on testing sequences to demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM PUBLISHER]