Weakly-supervised image segmentation is a challenging problem with multidisciplinary applications in multimedia content analysis and beyond. It aims to segment an by leveraging its image-level semantics (i.e., tags). This paper presents weakly-supervised algorithm that learns the distribution of spatially structural superpixel sets from labels. More specifically, we first extract graphlets given image, which are small-sized graphs consisting superpixels encapsulating their spatial structure....

Extracting discriminative and robust features from video sequences is the first most critical step in human action recognition. In this paper, instead of using handcrafted features, we automatically learn spatio-temporal motion for This achieved via an evolutionary method, i.e., genetic programming (GP), which evolves feature descriptor on a population primitive 3D operators (e.g., 3D-Gabor wavelet). way, scale shift invariant can be effectively extracted both color optical flow sequences....

The development of detection methods for oriented object remains a challenging task. A considerable obstacle is the wide variation in shape (e.g., aspect ratio) objects. Sample selection general has been widely studied as it plays crucial role performance method and achieved great progress. However, existing sample strategies still overlook some issues: (1) most them ignore information; (2) they do not make potential distinction between selected positive samples; (3) can only be applied to...

Zero-shot learning (ZSL) aims to recognize novel classes by transferring semantic knowledge from seen unseen ones. Semantic is learned attribute descriptions shared between different classes, which are strong prior for localization of object representing discriminative region features enabling significant visual-semantic interaction. Although few attention-based models have attempted learn such in a single image, the transferability and visual typically neglected. In this paper, we propose...

Typical representations for arbitrary-oriented object detection tasks include the oriented bounding box (OBB), quadrilateral (QBB), and point set (PointSet). Each representation encounters problems that correspond to its characteristics, such as boundary discontinuity, square-like problems, ambiguity, isolated points, which lead inaccurate detection. Although many effective strategies have been proposed various representations, there is still no unified solution. Current methods based on...

In view-based 3D object retrieval and recognition, each is described by multiple views. A central problem how to estimate the distance between two objects. Most conventional methods integrate distances of view pairs across objects as an estimation their distance. this paper, we propose a discriminative probabilistic modeling approach. It builds models for based on distribution its views, defined upper bound Kullback-Leibler divergence corresponding models. recognition accomplished measures....

Hand posture recognition (HPR) is quite a challenging task, due to both the difficulty in detecting and tracking hands with normal cameras limitations of traditional manually selected features. In this article, we propose two-stage HPR system for Sign Language Recognition using Kinect sensor. first stage, an effective algorithm implement hand detection tracking. The incorporates color depth information, without specific requirements on uniform-colored or stable background. It can handle...

Unbalance and reactive power currents of electric railway will result in poor quality three-phase industrial grid. A co-phase traction supply system is introduced this paper to solve these problems by adopting single-phase ac-dc-ac converter. The converter connected between two secondary windings transformer with abilities active transmission compensation. Because the conversion, can utilize feeding connection scheme instead traditional two-phase scheme. An actual 10 MVA/27.5 kV...

To address the hazy weather image degradation problem, we propose a single dehazing method based on physical model and brightness components of by using multi-scale retinex with color restoration algorithm. The overall process involves three components, including atmospheric light value calculation, transmission map estimation, recovery scene radiance. Our contribution is that novel algorithm to dehaze calculating computing while considering dynamic range image. Experimental results show our...

Human activity analysis in videos has increasingly attracted attention computer vision research with the massive number of now accessible online. Although many recognition algorithms have been reported recently, representation is challenging. Recently, manifold regularized sparse coding obtained promising performance action recognition, because it simultaneously learns and preserves structure. In this paper, we propose a generalized version Laplacian for human called p-Laplacian (pLSC). The...

Multi-class object detection in remote sensing images plays an important role many applications but remains a challenging task because of scale imbalance and arbitrary orientations the objects with extreme aspect ratios. In this paper, Asymmetric Feature Pyramid Network (AFPN), Dynamic Alignment (DFA) module, Area-IoU regression loss are proposed on basis one-stage cascaded method for multi-class images. The designed asymmetric convolutional block is embedded into AFPN handling ratios...

Although deep learning-based surface defect detection approaches have performed remarkably well in recent years, the complicated shapes and large size differences of defects still pose enormous challenges for most existing methods. To address these issues, we propose a novel method joining spatial deformable convolution dense feature pyramid, named SDDF-Net. First, construct convolution-based extraction network, which uses dynamic convolutional kernel with information to increase capability...

Although recurrent network-based optical flow estimation methods have shown great success in recent years, most of these difficulty handling large displacements and occlusions because the existing networks are usually restricted to coarse-resolution single-scale models while ignoring multiscale features brought by hierarchical concepts previous coarse-to-fine approaches. In this paper, we propose an adaptive-aware correlation network for estimation, named ACR-Net, which preserves fine motion...

In order to address the resource service optimal-selection (RSOS) and composition problem in manufacturing grid (MGrid) system provide high-quality users, an MGrid RSOS framework (MGrid-RSOSCF) is investigated this study. The process of divided into following five steps MGrid-RSOSCF: (1) decomposing submitted task several subtasks (i.e. single requested task) if a multiple task; (2) searching out qualified for each decomposed subtask generating corresponding candidate set; (3) retrieving,...

Binary hashing has been widely used for efficient similarity search due to its query and storage efficiency. In most existing binary methods, the high-dimensional data are embedded into Hamming space distance or of two points approximated by between their codes. The calculation is efficient, however, in practice, there often lots results sharing same a query, which makes this measure ambiguous poses critical issue where ranking important. paper, we propose weighted algorithm (WhRank) rank...

A large group of dictionary learning algorithms focus on adaptive sparse representation data. Almost all them fix the number atoms in iterations and use unfeasible schemes to update process. It's difficult, therefore, for train a from Big Data. new algorithm is proposed here by extending classical K-SVD method. In method, when each batch data samples added training process, are selectively introduced into dictionary. Furthermore, only small as subspace controls current orthogonal matching...

Notice of Violation IEEE Publication Principles<br><br> "Recent Advances in 3D Object Detection the Era Deep Neural Networks: A Survey,"<br> by M. Rahman, Y. Tan, J. Xue and K. Lu,<br> Transactions on Image Processing, vol. 29, 2020, pp. 2947-2962<br><br> After careful considered review content authorship this paper a duly constituted expert committee, has been found to be violation IEEE's Principles.<br><br> This contains portions text from cited below that were paraphrased without...