Search

Vehicle detection in satellite images has attracted extensive research interest with widespreading application potentials. The main challenge lies in the difficulty of labeling sufficient training instances (vehicle rectangles) across all resolutions and imaging conditions of satellite images, which degenerates the performance of vehicle detectors trained correspondingly. To tackle this challenge, in this paper we propose an intelligent and labor-light scheme for large-scale training of vehicle detectors. Our scheme only requires region-level group annotation, i.e. whether this region contains vehicle(s) or not, without explicitly labeling the bounding boxes of vehicles. To this end, a novel weakly supervised, multi-instance learning algorithm is designed to learn instance-wise vehicle detectors from such “weak labels”. In particular, a density estimator is firstly adopted to estimate the density map of vehicle instances from the positive regions. Then, a multi-instance SVM is trained to classify and locate vehicle instances from this map. We have carried out extensive experiments on a large-scale satellite image collection that contains various resolutions and imaging conditions. We have demonstrated that the proposed scheme has achieved superior performance by comparing to a set of state-of-the-art and alternative approaches. [ABSTRACT FROM AUTHOR]

Very recently, vehicle detection in satellite images has become an emerging research topic with various applications ranging from military to commercial systems. However, it retains as an open problem, mainly due to the complex variations in imaging conditions, object intra-class changes, as well as due to its low-resolution. Coming with the rapid advances in deep learning for feature representation, in this paper we investigate the possibility to exploit deep neural features towards robust vehicle detection. In addition, along with the rapid growth in the data volume, new classification methodology is also demanded to explicitly handle the intra-class variations. In this paper, we propose a vehicle detection framework, which combines Deep Convolutional Neural Network (DNN) based feature learning with Exemplar-SVMs (E-SVMS) based, robust instance classifier to achieve robust vehicle detection in satellite images. In particular, we adopt DNN to learn discriminative image features, which has a high learning capacity. In our practice, the leverage of DNN has achieve significant performance boost by comparing to a serial of handcraft designed features. In addition, we adopt E-SVMs based robust classifier to further improve the classification robustness, which can be considered as an instance-specific metric learning scheme. By conducting extensive experiments with comparisons to a serial of state-of-the-art and alternative works, we further show that the combination of both schemes can benefit from each other to jointly improve the detection accuracy and effectiveness. [ABSTRACT FROM AUTHOR]

The estimations of depth and regional semantics from a single image have traditionally been considered as two separated problems. In this paper, we argue that these two tasks provide complementary information, which therefore can be performed jointly to reinforce individual tasks in terms of both accuracy and speed. In particular, we propose an Elastic Conditional Random Field (E-CRF) deployed upon superpixel segmentations, which models the interdependency between depth and semantics to refine each other in an iterative manner. Differing from the traditional CRFs, E-CRF makes edges elastically hidden/emergent during inference to conduct fast Loopy Belief Propagation, while explicitly modeling the depth-label interdependency to achieve high inference accuracy. Moreover, the Structured Support Vector Machine (SSVM) is further introduced to drastically speed up the inference. We have conducted extensive evaluations on both Make3D and NYU benchmark datasets, which demonstrated that our E-CRF method significantly outperforms state-of-the-art techniques in terms of precision, while significantly accelerating the inference speed (2–3 orders of magnitude). [ABSTRACT FROM AUTHOR]

With the recent advances of low-cost depth sensors, there are increasing researches which focus on applications using such devices. A critical issue involved is the processing of the low-quality depth images. This paper presents an example-based method for filling the holes in the depth images. The method utilizes the redundant geometrical information contained in the repeated occurrences of some similar objects, so as to mutually infill these incomplete objects. We first introduce a user-assisted object segmentation to select the objects of interest. Subsequently, the method employs an object recognition procedure to detect the occurrences of the selected objects. Given these detected object instances, we complete each object using the detected partial range data in a registration way. Finally, the completed objects are synthesized back to the original depth image, thus to infill the holes. We demonstrate the effectiveness of the proposed method and the therein algorithms by experiments on several real scenes, which contains repeated objects either of one pattern or multiple patterns. [ABSTRACT FROM AUTHOR]

With the speedy development of social media, more and more multimedia data are generated by users with tags associated. The tag information provides the extra cue to link multimedia data in addition to the multimedia content itself. However, the manually added tags are always with noise and not correct enough. Moreover, the semantically similar tags exist massively but cannot be accounted for well. This paper proposes a new algorithm to robustly combine multimedia content and associated tags by mining the latent semantic which takes into account the semantically similar tags. The l 2 , 1 norm is proposed to employ in latent semantic indexing for a more robust latent space, and a word-to-vector based clustering method is proposed to address the massive tags with similar meaning. The experiments on extensive data demonstrate the proposed method. Compared to the existing latent semantic based methods, the algorithm proposed a more robust model to deal with noise. [ABSTRACT FROM AUTHOR]

Depth is a rich source of information and has been successfully utilized in numerous computer vision applications. However, it is often ignored in object tracking. In this paper, in contrast to traditional 2D image-based tracking method, we propose a novel 3D object tracking method from a moving binocular camera. To effectively handle the deformable targets, a target is first represented by a local patch-based appearance model. Then, to handle the partial occlusions, we design a simple yet effective scheme to detect and recovery occlusions using depth information obtained from a moving binocular camera. Therefore, the proposed method can simultaneous capture target appearance changes and alleviate the drifting problem. The experimental results demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

This paper describes an easy-to-use system to estimate the shape of a human body and his/her clothes. The system uses a Kinect to capture the human׳s RGB and depth information from different views. Using the depth data, a non-rigid deformation method is devised to compensate motions between different views, thus to align and complete the dressed shape. Given the reconstructed dressed shape, the skin regions are recognized by a skin classifier from the RGB images, and these skin regions are taken as a tight constraints for the body estimation. Subsequently, the body shape is estimated from the skin regions of the dressed shape by leveraging a statistical model of human body. After the body estimation, the body shape is non-rigidly deformed to fit the dressed shape, so as to extract the cloth field of the dressed shape. We demonstrate our system and the therein algorithms by several experiments. The results show the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Abstract: In recent years, extensive research efforts have been dedicated to automatic news content analysis. In this paper, we propose a novel algorithm for anchorperson detection in news video sequences. In this method, the raw news videos are firstly split into shots by a four-threshold method, and the key frames are extracted from each shot. After that, the anchorperson detection is conducted from these key frames by using a clustering-based method based on a statistical distance of Pearson's correlation coefficient. To evaluate the effectiveness of the proposed method, we have conducted experiments on 10 news sequences. In these experiments, the proposed scheme achieves a recall of 0.96 and a precision of 0.97 for anchorperson detection. [Copyright &y& Elsevier]

Abstract: Visual codebook serves as a fundamental component in many state-of-the-art visual search and object recognition systems. While most existing codebooks are built based solely on unsupervised patch quantization, there are few works exploited image labels to supervise its construction. The key challenge lies in the following: image labels are global, but patch supervision should be local. Such imbalanced supervision is beyond the scope of most existing supervised codebooks [9,10,12–15,29]. In this paper, we propose a weakly supervised codebook learning framework, which integrates image labels to supervise codebook building with two steps: the Label Propagation step propagates image labels into local patches by multiple instance learning and instance selection [20,21]. The Graph Quantization step integrates patch labels to build codebook using Mean Shift. Both steps are co-optimized in an Expectation Maximization framework: the E-phase selects the best patches that minimize the semantic distortions in quantization to propagate image labels; while the M-phase groups similar patches with related labels (modeled by WordNet [18]), which minimizes the visual distortions in quantization. In quantitative experiments, our codebook outperforms state-of-the-art unsupervised and supervised codebooks [1,10,11,25,29] using benchmark datasets. [Copyright &y& Elsevier]

Abstract: In this paper, we present a spatiotemporal co-location video pattern mining approach with application to robust action retrieval in YouTube videos. First, we introduce an attention shift scheme to detect and partition the focused human actions from YouTube videos, which is based upon the visual saliency [13] modeling together with both the face [35] and body [32] detectors. From the segmented spatiotemporal human action regions, we extract 3D-SIFT [17] detector. Then, we quantize all detected interest points from the reference YouTube videos into a vocabulary, based on which assign each individual interest point with a word identity. An APrior based frequent itemset mining scheme is then deployed over the spatiotemporal co-located words to discover co-location video patterns. Finally, we fuse both visual words and patterns and leverage a boosting based feature selection to output the final action descriptors, which incorporates the ranking distortion of the conjunctive queries into the boosting objective. We carried out quantitative evaluations over both KTH human motion benchmark [26], as well as over 60-hour YouTube videos, with comparisons to the state-of-the-arts. [Copyright &y& Elsevier]

The reversible watermarking technique is suitable for vector maps due to its reversibility after watermark extraction. In this paper, a novel reversible watermarking scheme based on the idea of nonlinear scrambling is proposed. It begins with feature point extraction. To avoid the high-precision vector data being illegally used by unauthorized users, the algorithm nonlinearly scrambles the relative position of feature points. Then based on the proposed reversible embedding, both scrambled feature points and nonfeature points are taken as cover data, the coordinates of which are modified to embed both watermark data and feature point identification data. Finally, combined with the scrambling secret key, the original vector data can be exactly recovered with watermark extraction. Comprehensive experimental results validate that the scheme could effectively prevent the high-precision vector data from being illegally used with maintaining the basic shape of each polyline, simultaneously. [ABSTRACT FROM AUTHOR]

We attempt to reduce the computational costs in vision transformers (ViTs), which increase quadratically in the token number. We present a novel training paradigm that trains only one ViT model at a time, but is capable of providing improved image recognition performance with various computational costs. Here, the trained ViT model, termed super vision transformer (SuperViT), is empowered with the versatile ability to solve incoming patches of multiple sizes as well as preserve informative tokens with multiple keeping rates (the ratio of keeping tokens) to achieve good hardware efficiency for inference, given that the available hardware resources often change from time to time. Experimental results on ImageNet demonstrate that our SuperViT can considerably reduce the computational costs of ViT models with even performance increase. For example, we reduce 2 × FLOPs of DeiT-S while increasing the Top-1 accuracy by 0.2% and 0.7% for 1.5 × reduction. Also, our SuperViT significantly outperforms existing studies on efficient vision transformers. For example, when consuming the same amount of FLOPs, our SuperViT surpasses the recent state-of-the-art EViT by 1.1% when using DeiT-S as their backbones. The project of this work is made publicly available at https://github.com/lmbxmu/SuperViT. [ABSTRACT FROM AUTHOR]

The person re-identification is an active research branch of the computer vision and attracts many researchers study on it. However, because of the variance in viewpoints, illumination, pedestrians' pose and some of other factors, building a robust person re-identification algorithm in open-world is rather challenging. In this paper, we propose two algorithms, indiscriminative patches trim strategy and multi-instance multi-label learning method, to fix the person re-identification in closed short-term surveillance network. There are two main contributions of this paper: first, we define the discriminative region in a person's image. Correspondingly, an adapted canopy-kmeans algorithm is proposed to evaluate the discriminability of image patches which form the discriminative regions. Besides, two strategies, in local image and in global gallery dataset, are proposed to filter out distractions. Finally, the true discriminative patches are employed to compute the similarity between two images, and then used to re-identify pedestrians. Second, we first introduce the multi-instance multi-label learning methods into re-identification, and hence we propose a framework of solving person re-identification by applying MIMLL. We employ two MIMLL methods, the MIMLBOOST and the MIMLSVM, to detect attributes in each image, and those detected attributes contribute to recognize different pedestrians. Experiments on two benchmark datasets show the competitive performance of the two proposed algorithms. [ABSTRACT FROM AUTHOR]

Nowadays by employing many machine learning and pattern classification methods in object classification, the view-based 3D object classification, an emerging research topic, becomes a major research focus. However, most existing researches focus on only a single modality of image features for the object classification, although recent studies have shown that different kinds of features may provide complementary information for 3D object classification. In this paper, we propose the multimodal support vector machine to combine three modalities of image features, i.e., Sift descriptor, Outline Fourier transform descriptor, and Zernike Moments descriptor to discriminate the multiple classes of object, where each kernel corresponds to each modality. In this way, not only the independence of each modality but also the interrelation between them are both taken into considered. And we further employ multi-task feature selection via the l 2 -norm regularization after feature extraction to improve the performance of final classification. The experiments conducted in ETH-80 image set demonstrate the effectiveness and superiority of our method. [ABSTRACT FROM AUTHOR]

In this paper, we present a novel aperiodic directional structure measurement (ADSM) toward road network extraction. Based on the observations from Cognitive Psychology regarding the aperiodicity and local directionality, ADSM can well characterize roadlike structures independent of the spectral character and contrast. By exploiting such measurement as guidance, we construct a mask to denote potential road regions. Then, by combining with some common morphology operators, our approach is able to provide robust road centerlines efficiently. We evaluate our approach with data from various satellite sensors and make comprehensive comparisons with previous state-of-the-art methods. Experimental results demonstrate the merit using our ADSM as a metric to identify potential road structures, as well as the effectiveness and efficiency of our road network extraction system. [ABSTRACT FROM AUTHOR]

Coming with the big data era, the filtering of uninformative data becomes emerging. To this end, ranking the high dimensionality features plays an important role. However, most of the state-of-art methods focus on improving the classification accuracy while the stability of the dimensionality reduction is simply ignored. In this paper, we proposed a Max-Relevance-Max-Distance (MRMD) feature ranking method, which balances accuracy and stability of feature ranking and prediction task. In order to prove the effectiveness on big data, we tested our method on two different datasets. The first one is image classification, which is a benchmark dataset with high dimensionality, while the second one is protein–protein interaction prediction data, which comes from our previous private research and has massive instances. Experiments prove that our method maintained the accuracy together with the stability on both two big datasets. Moreover, our method runs faster than other filtering and wrapping methods, such as mRMR and Information Gain. [ABSTRACT FROM AUTHOR]

Despite the exciting performance, Transformer is criticized for its excessive parameters and computation cost. However, compressing Transformer remains as an open problem due to its internal complexity of the layer designs, i.e., Multi-Head Attention (MHA) and Feed-Forward Network (FFN). To address this issue, we introduce Group-wise Transformation towards a universal yet lightweight Transformer for vision-and-language tasks, termed as LW-Transformer. LW-Transformer applies Group-wise Transformation to reduce both the parameters and computations of Transformer, while also preserving its two main properties, i.e., the efficient attention modeling on diverse subspaces of MHA, and the expanding-scaling feature transformation of FFN. We apply LW-Transformer to a set of Transformer-based networks, and quantitatively measure them on three vision-and-language tasks and six benchmark datasets. Experimental results show that while saving a large number of parameters and computations, LW-Transformer achieves very competitive performance against the original Transformer networks for vision-and-language tasks. To examine the generalization ability, we apply LW-Transformer to the task of image classification, and build its network based on a recently proposed image Transformer called Swin-Transformer, where the effectiveness can be also confirmed. [ABSTRACT FROM AUTHOR]

With the blooming of online social media applications, Community Question Answering (CQA) services have become one of the most important online resources for information and knowledge seekers. A large number of high quality question and answer pairs have been accumulated, which allow users to not only share their knowledge with others, but also interact with each other. Accordingly, volumes of efforts have been taken to explore the questions and answers retrieval in CQA services so as to help users to finding the similar questions or the right answers. However, to our knowledge, less attention has been paid so far to question popularity in CQA. Question popularity can reflect the attention and interest of users. Hence, predicting question popularity can better capture the users’ interest so as to improve the users’ experience. Meanwhile, it can also promote the development of the community. In this paper, we investigate the problem of predicting question popularity in CQA. We first explore the factors that have impact on question popularity by employing statistical analysis. We then propose a supervised machine learning approach to model these factors for question popularity prediction. The experimental results show that our proposed approach can effectively distinguish the popular questions from unpopular ones in the Yahoo! Answers question and answer repository. [ABSTRACT FROM AUTHOR]

As an important pre-processing stage in many machine learning and pattern recognition domains, feature selection deems to identify the most discriminate features for a compact data representation. As typical feature selection methods, Lasso and its variants using the l 1 -norm based regularization have received much attention in recent years. However, most of existing l 1 -norm based sparse feature selection methods ignore the structure information of data or only consider the pairwise relationships among samples. In this paper, we propose a hypergraph regularized sparse feature learning method, where the high-order relationships among samples are modeled and incorporated into the learning process. Specifically, we first construct a hypergraph with multiple hyperedges to capture the high-order relationships among samples, followed by the computation of a hypergraph Laplacian matrix. Then, we propose a hypergraph regularization term, and a hypergraph regularized Lasso model. We conduct a series of experiments on a number of data sets from UCI machine learning repository, and two real-world neuroimaging based classification tasks. Experimental results demonstrate that the proposed method achieves promising classification results, compared with several well known feature selection approaches. [ABSTRACT FROM AUTHOR]

In this paper, we propose deep architecture to dynamically learn the most discriminative features from data for both single-cell and object tracking in computational biology and computer vision. Firstly, the discriminative features are automatically learned via a convolutional deep belief network (CDBN). Secondly, we design a simple yet effective method to transfer features learned from CDBNs on the source tasks for generic purpose to the object tracking tasks using only limited amount of training data. Finally, to alleviate the tracker drifting problem caused by model updating, we jointly consider three different types of positive samples. Extensive experiments validate the robustness and effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Tracking individual-cell/object over time is important in understanding drug treatment effects on cancer cells and video surveillance. A fundamental problem of individual-cell/object tracking is to simultaneously address the cell/object appearance variations caused by intrinsic and extrinsic factors. In this paper, inspired by the architecture of deep learning, we propose a robust feature learning method for constructing discriminative appearance models without large-scale pretraining. Specifically, in the initial frames, an unsupervised method is firstly used to learn the abstract feature of a target by exploiting both classic principal component analysis (PCA) algorithms with recent deep learning representation architectures. We use learned PCA eigenvectors as filters and develop a novel algorithm to represent a target by composing of a PCA-based filter bank layer, a nonlinear layer, and a patch-based pooling layer, respectively. Then, based on the feature representation, a neural network with one hidden layer is trained in a supervised mode to construct a discriminative appearance model. Finally, to alleviate the tracker drifting problem, a sample update scheme is carefully designed to keep track of the most representative and diverse samples during tracking. We test the proposed tracking method on two standard individual cell/object tracking benchmarks to show our tracker's state-of-the-art performance. [ABSTRACT FROM AUTHOR]

Consulting through message-updating in social network is regarded as a popular way of information seeking. However, most questions cannot receive answers or suggestions timely, and some questions even fail to get replies. Thus, identifying microblogs that contain questions (we call them 'question microblog') and recommending answers automatically are meaningful. We divide this problem into two submodules: question identification and answer recommendation. To the best of our knowledge, few attempts have been made to identify questions in microblogs due to standard features such as 5W1H (How, What, Where, When, Who, Why) are likely to be absent. The following challenging problem is how to provide users a relevant, credible, diversified and personalized answer after a microblog is recognized as a question. In this paper, we investigate the feasibility of integrating standard features and contextual features extracted from auxiliary resources and recommend a reasonable answer using collaborative filtering. Empirical results on Sina Microblog dataset demonstrate the efficacy and effectiveness of our method. [ABSTRACT FROM AUTHOR]

This paper presents a study of vehicle detection from high-resolution aerial images. In this paper, a superpixel segmentation method designed for aerial images is proposed to control the segmentation with a low breakage rate. To make the training and detection more efficient, we extract meaningful patches based on the centers of the segmented superpixels. After the segmentation, through a training sample selection iteration strategy that is based on the sparse representation, we obtain a complete and small training subset from the original entire training set. With the selected training subset, we obtain a dictionary with high discrimination ability for vehicle detection. During training and detection, the grids of histogram of oriented gradient descriptor are used for feature extraction. To further improve the training and detection efficiency, a method is proposed for the defined main direction estimation of each patch. By rotating each patch to its main direction, we give the patches consistent directions. Comprehensive analyses and comparisons on two data sets illustrate the satisfactory performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Abstract: Single and cross-camera multiple person tracking in unconstrained condition is an extremely challenging task in computer vision. Facing the main difficulties caused by the existence of occlusion in single-camera scenario and the occurrence of transition in cross-camera scenario, we propose a unified framework formulated in graph matching with affinity constraints for both single and cross-camera tracking tasks. To our knowledge, our work is the first to unify two kinds of tracking problems with the same framework by graph matching. The proposed method consists of two steps, tracklet generation and tracklet association. First, we implement the modified part-based human detector and the Tracking-Modeling-Detection (TMD) method for tracklet generation. Then we propose to associate tracklets by graph matching which is mathematically formulated into the Rayleigh Quotients Maximization. The comparison experiments show that the proposed method can produce the competing results with the state-of-the-art methods. [Copyright &y& Elsevier]

With the blooming of Web 2.0, Community Question Answering (CQA) services such as Yahoo! Answers (http://answers.yahoo.com), WikiAnswer (http://wiki.answers.com), and Baidu Zhidao (http://zhidao.baidu.com), etc., have emerged as alternatives for knowledge and information acquisition. Over time, a large number of question and answer (Q&A) pairs with high quality devoted by human intelligence have been accumulated as a comprehensive knowledge base. Unlike the search engines, which return long lists of results, searching in the CQA services can obtain the correct answers to the question queries by automatically finding similar questions that have already been answered by other users. Hence, it greatly improves the efficiency of the online information retrieval. However, given a question query, finding the similar and well-answered questions is a non-trivial task. The main challenge is the word mismatch between question query (query) and candidate question for retrieval (question). To investigate this problem, in this study, we capture the word semantic similarity between query and question by introducing the topic modeling approach. We then propose an unsupervised machine-learning approach to finding similar questions on CQA Q&A archives. The experimental results show that our proposed approach significantly outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]