Object detection in point cloud data is one of the key components computer vision systems, especially for autonomous driving applications. In this work, we present Voxel-Feature Pyramid Network, a novel one-stage 3D object detector that utilizes raw from LIDAR sensors only. The core framework consists an encoder network and corresponding decoder followed by region proposal network. Encoder extracts fuses multi-scale voxel information bottom-up manner, whereas multiple feature maps various...

Drug repositioning has shorter developmental time, lower cost and less safety risk than traditional drug development process. The current study aims to repurpose marketed drugs clinical candidates for new indications in diabetes treatment by mining 'omics' data. We analyzed data from genome wide association studies (GWAS), proteomics metabolomics revealed a total of 992 proteins as potential anti-diabetic targets human. Information on the that target these was retrieved Therapeutic Target...

We report XLINK, a multi-path QUIC video transport solution with experiments in Taobao short videos. XLINK is designed to meet two operational challenges at the same time: (1) Optimized user-perceived quality of experience (QoE) terms robustness, smoothness, responsiveness, and mobility (2) Minimized cost overhead for service providers (typically CDNs). The core take opportunity as user-space protocol directly capture QoE intent control scheduling management. overcome major hurdles such...

This paper studied generating natural languages at particular contexts or situations. We proposed two novel approaches which encode the into a continuous semantic representation and then decode text sequences with recurrent neural networks. During decoding, context information are attended through gating mechanism, addressing problem of long-range dependency caused by lengthy sequences. evaluate effectiveness on user review data, in rich available informative contexts, sentiments products,...

Real-time and high-performance 3D object detection is an attractive research direction in autonomous driving. Recent studies prefer point based or voxel convolution for achieving high performance. However, these methods suffer from the unsatisfied efficiency complex customized convolution, making them unsuitable applications with real-time requirements. In this paper, we present efficient effective framework, named RangeIoUDet that uses range image as input. Benefiting dense representation...

We present RangeRCNN, a novel and effective 3D object detection framework based on the range image representation. Most existing methods are voxel-based or point-based. Though several optimizations have been introduced to ease sparsity issue speed up running time, two representations still computationally inefficient. Compared them, representation is dense compact which can exploit powerful 2D convolution. Even so, not preferred in due scale variation occlusion. In this paper, we utilize...

This paper proposes a novel inertial-aided localization approach by fusing information from multiple inertial measurement units (IMUs) and exteroceptive sensors. IMU is low-cost motion sensor which provides measurements on angular velocity gravity compensated linear acceleration of moving platform, widely used in modern systems. To date, most existing methods exploit only one single IMU. While the single-IMU yields acceptable accuracy robustness for different use cases, overall performance...

Abstract Mountain railway alignment design is an important but complex civil engineering problem. To overcome the drastically undulating terrain, long tunnels and high bridges are major structures used along a mountain railway, which poses great challenges for construction. Unfortunately, despite being studied many years, crucial construction factors of have received slight attention in optimization. In this paper, first time, layout large‐scale auxiliary projects (LACPs), including tunnel...

In this work, we propose a novel method for performing inertial aided navigation, by using deep neural net-works (DNNs). To date, most DNN navigation methods focus on the task of odometry, taking gyroscope and accelerometer readings as input regressing integrated IMU poses (i.e., position orientation). While design has been successfully applied number applications, it is not theoretical performance guarantee unless patterned motion involved. This inevitably leads to significantly reduced...

Camera and LIDAR are both important sensor modalities for real-world applications, especially autonomous driving. The sensors provide complementary information make fusion possible. However, the progress of early-fusion is very slow due to limitations viewpoint misalignment, feature misalignment data volume alignment, so that its performance also low. In this work, we propose a novel pipeline: an method range image RGB enhance 3D object detection. It takes full advantage LIDAR's view, point...

We present GSO-Simulcast, a new architecture designed for large-scale multi-party video-conferencing systems. GSO-Simulcast is currently deployed at full-scale in Alibaba's Dingtalk video conferencing that serves more than 500 million users. It marks fundamental shift from today's Simulcast, where media server locally decides how to switch and forward streams based on fragmented network view. Instead, globally orchestrates the publishing, subscribing, as well resolution bitrate of each...

In this paper we propose a new method to reduce noise in digital images. The is based on the bilateral filter. filter nonlinear that does spatial averaging without smoothing edges. aspect of very crucial; has been shown work better than wavelet thresholding some recent papers. proposed improves through decomposing signal into its frequency components. way, different components can be eliminated. Experimental results with both simulated and real images are given. addition method, also provide...

This paper proposes a highly efficient approach for power disturbance waveform (PDW) compression in the view of Heisenberg uncertainty. The key idea is to represent each signal component PDW using as few nonzero coefficients possible by uncertainty principle restriction. PDWs are projected union bases (UB), and can be represented very sparsely. UB decomposition solved orthogonal matching pursuit. features cross correlation subbases guarantee with high ratio recovered accuracy. With various...

In this paper, we present a spatially adaptive method to reduce compression artifacts observed in block discrete cosine transform (DCT) based image/video standards. The is on the bilateral filter, which very effective denoising images without smoothing edges. When applied artifacts, parameters of filter should be chosen carefully have good performance. To avoid over-smoothing texture regions and effectively eliminate blocking ringing boundary discontinuities are first detected; these then...

Photo Response Non-Uniformity (PRNU) noise-based source camera attribution is a popular digital forensic method. In this method, fingerprint computed from set of known images the matched against extracted noise an anonymous questionable image to find out if had taken image. The possibility privacy leak, however, one main concerns PRNU-based Using (or noise), adversary can identify owner by matching with images) crawled social media account. article, we address concern encrypting both and...

Neutrosophy studies the origin, nature, scope of neutralities, and their interactions with different ideational spectra. It is a new philosophy that extends fuzzy logic basis neutrosophic logic, probability, set theory, statistics. Because world full indeterminacy, imperfection knowledge human receives/observes from external also causes imprecision. introduces concept , which representation indeterminacy. However, this theory mostly discussed in physiology mathematics. Thus, applications...

The 2D to 3D conversion technique plays a crucial role in the development and promotion of three-dimensional television (3DTV) for it can provide adequate supply high-quality program material. In this paper, novel automatic method using multi-depth cues is presented. depth used our system, which will be integrated into one map according types scenes, include perspective geometry, defocus, visual saliency adaptive models. After maps are extracted, original image or video converted...

Early and highly precise detection is essential for delaying the progression of coronary artery disease (CAD). Previous methods primarily based on single-modal data inherently lack sufficient information that compromises precision. This paper proposes a novel multi-modal learning method aimed to enhance CAD by integrating ECG, PCG, coupling signals. A signal initially generated operating deconvolution ECG PCG. Then, various entropy features are extracted from its signals, as well recurrence...

The bilateral filter is a nonlinear that does spatial averaging without smoothing edges; it has shown to be an effective image denoising technique in addition some other applications. There are two main contributions of this paper. First, we provide empirical study the optimal parameter selection for Second, present extension filter: multi-resolution filter, where filtering applied low-frequency subbands signal decomposed using orthogonal wavelet transform. Combined with thresholding, new...

This paper presents a novel performance-driven realtime facial animation system with 3ds Max and Kinect. The user performs in natural environment without any marker, the 3D video of is captured 2D key points are tracked improved Active Appearance Model, set Action Units derived iterative closest point algorithm. Based on two rough regions eyes, we performed extra operations to track blinks pupil motions. data transmitted as control parameters using Musical Instrument Device Interface...

Obstacle perception based on radar sensor has drawn wide attentions in autonomous driving due to robust performance and low cost. It is significant utilize fusion, e.g., camera information, further enhance the ability. Although much progress been made, we still observe two problems: First, spatial alignment among multi-modal data intractable when involving multiple sensors. Second, most existing works are object-level which inevitably information loss leading a degradation. To this end,...

Although the basic principle of Hough transformation can be described accurately in continuous spaces, its application is often conducted digitized ones. The discretization both a spatial image and related parameters will result positioning errors parameter domain that affect accumulation through which functions. This makes an important issue. Its resolution needs to carefully selected assure effective concentration accumulation. In this paper, effects digitization on latter are analyzed...

3D vehicle detection based on multi-modal fusion is an important task of many applications such as autonomous driving. Although significant progress has been made, we still observe two aspects that need to be further improvement: First, the specific gain camera images can bring seldom explored by previous works. Second, algorithms run slowly, which essential for with high real-time requirements(autonomous driving). To this end, propose end-to-end trainable single-stage feature adaptive...