The irregular domain and lack of ordering make it challenging to design deep neural networks for point cloud processing. This paper presents a novel framework named Point Cloud Transformer(PCT) learning. PCT is based on Transformer, which achieves huge success in natural language processing displays great potential image It inherently permutation invariant sequence points, making well-suited To better capture local context within the cloud, we enhance input embedding with support farthest...

Convolutional neural networks (CNNs) have made great breakthroughs in 2D computer vision. However, their irregular structure makes it hard to harness the potential of CNNs directly on meshes. A subdivision surface provides a hierarchical multi-resolution structure, which each face closed 2-manifold triangle mesh is exactly adjacent three faces. Motivated by these two observations, this paper presents SubdivNet, an innovative and versatile CNN framework for 3D meshes with Loop sequence...

Graph Neural Networks (GNNs) with attention have been successfully applied for learning visual feature matching. However, current methods learn complete graphs, resulting in a quadratic complexity the number of features. Motivated by prior observation that self- and cross- matrices converge to sparse representation, we propose ClusterGNN, an attentional GNN architecture which operates on clusters matching task. Using progressive clustering module adaptively divide keypoints into different...

We present a novel algorithm for semantic segmentation and labeling of 3D point clouds indoor scenes, where objects in can have significant variations complex configurations. Effective methods decomposing into semantically meaningful pieces are highly desirable object recognition, scene understanding, modeling, etc. However, existing based on low-level geometry tend to either under-segment or over-segment clouds. Our method takes fundamentally different approach, is achieved along with...

Compositing is one of the most important editing operations for images and videos. The process improving realism composite results often called harmonization. Previous approaches harmonization mainly focus on images. In this paper, we take step further to attack problem video Specifically, train a convolutional neural network in an adversarial way, exploiting pixel-wise disharmony discriminator achieve more realistic harmonized introducing temporal loss increase consistency between...

Abstract Dynamic contrast‐enhanced MRI of atherosclerotic vessels after contrast agent injection may provide unique information regarding lesion structure and vulnerability. The high‐resolution images necessary for viewing substructures, however, are often corrupted by patient motion low signal‐to‐noise ratios, making pixel‐level analyses difficult. This article presents a postprocessing method that enables analysis dynamic eliminating enhancing image quality. Noise correction performed...

Abstract Sparse view 3D reconstruction has attracted increasing attention with the development of neural implicit representation. Existing methods usually only make use 2D views, requiring a dense set input views for accurate reconstruction. In this paper, we show that can be achieved by incorporating geometric priors into Our method adopts signed distance function as representation, and learns generalizable surface model from sparse views. Specifically, build more effective feature volume...

Bounded round multiprover interactive proof systems (MIPs) are compared with unbounded (IPSs). It is shown that for any constant epsilon , language accepted by an IPS has a bounded round, two-prover MIP error probability resolving open problem of L. Fortnow et al. (1988). To obtain this result, it certain one-round simulates the computation can be executed many times in parallel to significantly reduce its error.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Compositing is one of the most important editing operations for images and videos. The process improving realism composite results often called harmonization. Previous approaches harmonization mainly focus on images. In this work, we take step further to attack problem video Specifically, train a convolutional neural network in an adversarial way, exploiting pixel-wise disharmony discriminator achieve more realistic harmonized introducing temporal loss increase consistency between...

Attention Graph Neural Network (GNNs) has proven to be highly effective in image feature matching tasks. However, this method encounters a challenge due its quadratic complexity the number of keypoints. To address issue, we propose sparse graph neural network, which reduces redundancy by grouping correlated points into same category. avoid impact misclassification during clustering process, introduce novel approach that utilizes centers as global tokens. We incorporate information message...

This paper presents a Semantic Positioning System (SPS) to enhance the accuracy of mobile device geo-localization in outdoor urban environments. Although traditional Global (GPS) can offer rough localization, it lacks necessary for applications such as Augmented Reality (AR). Our SPS integrates Geographic Information (GIS) data, GPS signals, and visual image information estimate 6 Degree-of-Freedom (DoF) pose through cross-view semantic matching. approach has excellent scalability support...

Summary Spectral analysis, as a useful method of researching non-stationary signals, plays an important role in the process geophysical interpretation. The goal spectral analysis is to obtain time-frequency spectrums which reveal frequency characteristics varying with time. This paper introduces adaptive shaping regularization analysis. In this method, Fourier series coefficients are regarded function time and can be computed directly by inverting truncated complex sinusoidal basis within...

Seismic signals are non-stational, and spectral decomposition is an important method for studying the properties of non-stationary signals. Conventional spectrum methods can not simultaneously have high time-frequency resolution, which satisfy requirement high-precision seismic data interpretation at present. Therefore, text proposes a new with resolution in time frequency domain named synchrosqueezing wavelet transform (SSWT). Since coefficients compressed rearranged only axis SSWT, it...