Feature matters for salient object detection. Existing methods mainly focus on designing a sophisticated structure to incorporate multi-level features and filter out cluttered features. We present Progressive Polishing Network (PFPN), simple yet effective framework progressively polish the be more accurate representative. By employing multiple Modules (FPMs) in recurrent manner, our approach is able detect objects with fine details without any post-processing. A FPM parallelly updates of...

Radiation treatment is considered an effective and the most common option for prostate cancer. The planning process requires accurate precise segmentation of organs at risk (OARs), which laborious time-consuming when contoured manually. Artificial intelligence (AI)-based auto-segmentation has potential to significantly accelerate radiation therapy process; however, accuracy needs be validated before its full clinical adoption.A commercial AI-based contouring model was trained provide...

In this letter, we study the uplink channel throughput performance of a proposed novel multiple-antenna hybrid-domain non-orthogonal multiple access (MA-HD-NOMA) scheme. This scheme combines conventional sparse code (SCMA) and power-domain NOMA (PD-NOMA) schemes in order to increase number users served as compared uses antennas at base station. To end, joint resource allocation problem for MA-HD-NOMA is formulated that maximizes sum rate entire system. For comprehensive comparison, problems...

Abstract Background Inter‐fractional and intra‐fractional positioning motion monitoring based on implanted fiducials through kilovoltage (kV) imaging is a cost‐effective approach to enhance intact prostate radiotherapy treatment accuracy. However, comprehensive studies for three‐dimensional (3D) position extraction the corrective action threshold are lacking. Purpose To develop verify fiducial system with 3D information triggered kV images taken during delivery, comprehensively evaluate...

In this paper, we consider distributed detection of sparse stochastic signals with quantized measurements. Assume that both the noise and dominant elements in follow generalized Gaussian (GG) distribution. Since communication bandwidth sensor networks is limited, local sensors send measurements instead analog data to fusion center. First, propose locally most powerful test (LMPT) detector based on for GG Then, quantizers are designed by maximizing efficacy proposed LMPT detector. particular,...

GPUs have been widely adopted to serve online deep learning-based services that stringent QoS requirements. However, emerging learning serving systems often result in long latency and low throughput of the inference requests damage user experience increase number required host an service. Our investigation shows poor batching operation lacking data transfer-computation overlap are root causes throughput. To this end, we propose Ebird, a system is comprised GPU-resident memory pool,...

The quantification of myelin water content in the brain can be obtained by multi-echo T2* weighted images (T2* WIs). To accelerate long acquisition, a novel tensor dictionary learning algorithm with low-rank and sparse regularization (TDLLS) is proposed to reconstruct WIs from undersampled data. explores local nonlocal similarity global temporal redundancy real imaginary parts complex relaxation signals. joint application constraints on dictionaries core coefficient tensors improves...

This paper employs an audit bit based mechanism to mitigate the effect of Byzantine attacks on distributed Bayesian detection systems. In this framework, optimal attacking strategy for strategic attackers is investigated traditional scheme (TAS) evaluate robustness system. We show that it possible a attacker degrade performance TAS system without bits. To enhance in presence attackers, we propose enhanced (EAS). The fusion rule proposed derived and evaluated via probability error Simulation...

Abstract Stroke correspondence construction is a precondition for vectorized 2D animation inbetweening and remains challenging problem. This paper introduces the FTP‐SC, fuzzy topology preserving stroke technique, which accurate provides user more effective control on result than previous matching approaches. The method employs two‐stage scheme to progressively establish between keyframes. In first stage, correspondences with high confidence are constructed by enforcing preservation of...

SPICE (Simulation Program with Integrated Circuit Emphasis) is a widely used circuit simulation framework for integrated designs. The basic skeleton of time domain was derived from the versions developed in UC Berkeley during 1970s. In this paper, we report most recent numerical integration methods to improve traditional schemes, which are based on linear multi-step and low order approximation differential equation system. Recently, matrix exponential algorithms being address long-term...

Abstract Stroke correspondence construction is a precondition for generating inbetween frames from set of key frames. In our case, each stroke in frame vector represented as Disk B‐Spline Curve (DBSC) which flexible and compact format. However, it not easy to construct correspondences between multiple DBSC strokes effectively because the following points: (1) with use shape descriptors, dimensionality feature space high; (2) number different usually large other (3) length corresponding can...

Aggressively provisioned data centers achieve great cost savings by over-committing the very expensive power distribution infrastructure. However, existing proposals for managing load demand in such a center are largely utilization-driven, overlooking power-related interferences among users. An important observation is that some tasks can impact budget management framework and disrupt normal operation taking away precious public capacity. This vulnerability exposes to new type of risk we...

In distributed detection systems, energy-efficient ordered transmission (EEOT) schemes are able to reduce the number of transmissions required make a final decision. this work, we investigate effect data falsification attacks on performance EEOT-based systems. We derive probability error for an system under attack and find upper bound (UB) expected Moreover, tighten UB by solving optimization problem via integer programming (IP). also obtain FC's optimal threshold which guarantees system....

We present a fast time-domain multiphysics simulation framework that combines full-wave electromagnetism (EM) and carrier transport in semiconductor devices (TCAD). The proposed features division of linear nonlinear components the EM-TCAD coupled system. former is extracted handled independently with high efficiency by matrix exponential approach assisted Krylov subspace method. latter treated ordinary Newton's method yet much sparser Jacobian leads to substantial speedup solving system...

This letter proposes a novel hybrid-domain (HD) non-orthogonal multiple access (NOMA) approach to support larger number of uplink users than the recently proposed code-domain NOMA approach, i.e., sparse code (SCMA). HD-NOMA combines and power-domain schemes by clustering in small path loss (strong) large (weak) groups. The two groups are decoded using successive interference cancellation while within group message passing algorithm. To further improve performance system, spectral-efficiency...

Computer-assisted auto coloring (CAAC) has great potential in terms of lowering production cost and saving time cel animation production. In this paper, a novel approach to automatically color the character line drawings based on region matching master frames is proposed. Firstly, some important attributes such as curve length, points, area etc. are investigated. Then, these attributes, detailed process presented followed by process. The results show that our can straightforwardly handle...

Abstract This paper proposes a Hierarchical Region Matching (HRM) approach for computer‐assisted auto coloring. The region‐level analysis in traditional 2D animation is expanded into several component levels with novel hierarchization method. With the hierarchy, various region matching algorithms can be applied from first/highest to last/lowest level. HRM improves accuracy and may deal errors caused by occlusion, thus making more robust, as verified results. Copyright © 2005 John Wiley &...

Emerging latency-critical (LC) services often have both CPU and GPU stages (e.g. DNN-assisted services) require short response latency. Co-locating best-effort (BE) applications on the side with LC service improves resource utilization. However, contention results in QoS violation of services. We therefore present CHARM, a collaborative host-accelerator management system. CHARM ensures required target services, while maximizing utilization host accelerator. is comprised BE-aware allocator,...

The ordered transmission based (OT-based) schemes reduce the number of transmissions needed in a distributed detection network without any loss probability error performance. In this paper, we investigate performance conventional OT-based system presence additive Byzantine attacks Gaussian shift mean problems. work, by launching attacks, attackers are able to alter order as well data for binary hypothesis testing problem. We also determine optimal attack strategy sensors. Furthermore,...

Content blocking is an important part of a performant, user-serving, privacy respecting web. Most content blockers build trust labels over URLs. While useful, this approach has well understood shortcomings. Attackers may avoid detection by changing URLs or domains, bundling unwanted code with benign code, inlining in pages. The common flaw existing approaches that they evaluate based on its delivery mechanism, not behavior. In work we address problem system for generating signatures the...