Visual grounding (VG) aims to locate a specific target in an image based on given language query. The discriminative information from context is important for distinguishing the other objects, particularly targets that have same category as others. However, most previous methods underestimate such information. Moreover, they are usually designed standard scene (without any novel object), which limits their generalization open-vocabulary scene. In this paper, we propose framework with...

microMVP is an affordable, portable, and open source micro-scale mobile robot platform designed for robotics research education. As a complete unique multi-vehicle enabled by 3D printing the maker culture, can be easily reproduced requires little maintenance: set of six micro vehicles, each measuring 8 × 5 6 cubic centimeters weighing under 100 grams, accompanying tracking fully assembled in two hours, all from readily available components. In this paper, we describe microMVP's hardware...

A percolation model for the critical current in inhomogeneous superconductors is introduced. The a network of randomly configured superconducting (concentration p) and normal 1-p) bonds on lattice. Each bond has ${i}_{c}$ above which it becomes Ohmic resistor. distribution regions solved using linearized Landau-Ginzburg equations wires as proposed by de Gennes. Kirchoff's laws. voltage-current relations are studied numerically two dimensions square lattice, comparisons made with recent...

Clustering is to group data so that the observations in same are more similar each other than those groups. k-means a popular clustering algorithm mining. Its objective optimize mean squared error (MSE). The traditional not suitable for applications where sizes of clusters need be balanced. Given n observations, our MSE under constraint evenly divided into k clusters. In this paper, we propose an iterative method task with balanced size constraints. Each iteration can split two steps, namely...

It is of vital importance to stitch the two images into a panorama in many computer vision applications motion detection and tracking virtual reality, panoramic photography, tours. To preserve more local details with few artifacts panoramas, this article presents an improved mesh-based joint optimization image stitching model. Since uniform vertices are usually used warps, we consider matched feature points as grid strengthen constraints on deformed vertices. Simultaneously, define energy...

Compositional models represent visual patterns as hierarchies of meaningful and reusable parts. They are attractive to vision modeling due their ability decompose complex into simpler ones resolve the lowlevel ambiguities in high-level image interpretations. However, current compositional separate structure part discovery from parameter estimation, which generally leads suboptimal learning fitting model. Moreover, commonly adopted latent structural is not scalable for deep architectures. To...

Clustering is one of the essential tools for data mining since it reveals natural structures unlabeled data. Many clustering algorithms have been proposed in last decades. However, few them are designed to adapt prior knowledge that available many real applications, such as sizes clusters. In this paper, we propose a novel iterative algorithm can impose constraints on Given an unordered set cluster size constraints, method minimizes mean squared error (MSE) while simultaneously considers...

This paper proposes a novel image large rotation and scale estimation method based on the Gabor filter pulse-coupled neural network (PCNN). First, features of template its rotated one are extracted by performing filter. Second, we present modified PCNN model to measure similarity between one. Finally, angle is calculated searching global minimum correlation coefficients. Besides estimation, also propose max-projection strategy. The feature projected along estimated angle, peak this...

Advances in robot locomotion algorithms, sensor technologies, planning frameworks and hardware designs have motivated the deployment of groups unstructured environments. One most common tasks is autonomous exploration. During multi-robot exploration task, it desirable for robots to coordinate order steer them different areas so as increase overall efficiency. Such coordination requires communication among peer robots. In some environments, such tunnels, natural caves or urban underground...

Abstract In this paper, a fast image registration algorithm based on SIFT is proposed, which slow in extracting feature points from high-resolution images. the original transformed into an pyramid to obtain sub-sampled with lower resolution. At same time, Laplace obtained. Feature of are extracted through algorithm, and then they fused image. Experimental results show that can quickly find obviously improve computational efficiency.

microMVP is an affordable, portable, and open source micro-scale mobile robot platform designed for robotics research education. As a complete unique multi-vehicle enabled by 3D printing the maker culture, can be easily reproduced requires little maintenance: set of six micro vehicles, each measuring $8\times 5\times 6$ cubic centimeters weighing under $100$ grams, accompanying tracking fully assembled in two hours, all from readily available components. In this paper, we describe microMVP's...

Abstract This paper proposes a new first-scan method for two-scan labeling algorithms. In the first scan, our proposed scans image lines three by with leaving line, and foreground pixels among each lines, assigns them provisional labels, finds resolves label equivalences them. Then, it processes from top to bottom one one, on line finding resolving between those immediately above below current line. Experimental results demonstrated that is more efficient than conventional label-equivalence-based

For the demand of recognising erotic image, a new method image based on multi-source information fusion is proposed. The makes use Gaussian skin model to implement detection naked area, combines Hough transform and Adaboost algorithm face, extracts features face color form eigenvector, uses SVM training obtain classifier. Adoptes extended Haar-like describe female breast, extract center surround feature, Calls separately two classifiers recognition, fuses recognition results with multiple...

Successive cancellation (SC) decoding algorithm and successive list (SCL) are two common methods for polar codes. The SC has low complexity, but in its process, the bits with small channel indices used to assist of large indices. Therefore, bit being decoded depends on that have already been which causes error propagation. On basis decoding, SCL overcomes problem propagation by widening path search width retaining multiple paths at same time process. However, complexity is higher than...

To make the moving object detection faster and more reliable, in this paper we present a novel method based on fast approximated SIFT descriptor. The main idea is to compute feature descriptor of key-point using integral histogram surrounding squared region. could be further used matching between two sequential frames image sequence. When involved calculating hundreds descriptors, profitable as it reduced computational cost, accelerated speed while still maintained fairly stable performance...

The training and running of neural network require large computational space memory space, which makes it difficult to deploy on a resource-constrained embedded sys-tems. To address this limitation, we introduce two-stage pipeline: Depth separable, local binary. Our method is divided into two steps. Firstly, deep separable convolution applied each input channel. Then, point-by-point the feature map obtained by filter. In second step, use binarization initialize filter corresponding channel...

This letter proposes an efficient joint resource allocation scheme for OFDM-based cooperative cognitive radio networks (CRNs) under various practical limitations.Compared with those traditional approaches, which guarantee the transmission rates of users, proposed ensures that users are maintained in proportion to predefined target rates.Numerical simulation shows can achieve a reasonable tradeoff between performance and computational complexity.