Generating free-viewpoint videos is critical for immersive VR/AR experience but recent neural advances still lack the editing ability to manipulate visual perception large dynamic scenes. To fill this gap, in paper we propose first approach editable photo-realistic video generation large-scale scenes using only sparse 16 cameras. The core of our a new layered representation, where each entity including environment itself formulated into space-time coherent radiance representation called...

Learning-based multi-view stereo (MVS) method heavily relies on feature matching, which requires distinctive and descriptive representations. An effective solution is to apply non-local aggregation, e.g., Transformer. Albeit useful, these techniques introduce heavy computation overheads for MVS. Each pixel densely attends the whole image. In contrast, we propose constrain nonlocal augmentation within a pair of lines: each point only corresponding epipolar lines. Our idea takes inspiration...

Simultaneous localization and mapping (SLAM) based on RGB-D cameras has been widely used for robot navigation in unknown environments. Most current SLAM methods are constrained by static environment assumptions perform poorly real-world dynamic scenarios. To improve the robustness performance of systems environments, this paper proposes a new method indoor scenes object detection. The presented improves ORB-SLAM3 framework. First, we designed an detection module YOLO v5 relied it to tracking...

Correspondence pruning aims to search consistent correspondences (inliers) from a set of putative correspondences. It is challenging because the disorganized spatial distribution numerous outliers, especially when are largely dominated by outliers. It's more ensure effectiveness while maintaining efficiency. In this paper, we propose an effective and efficient method for correspondence pruning. Inspired success attentive context in problems, first extend first-order then introduce idea...

Video stabilization refers to the problem of transforming a shaky video into visually pleasing one. The question how strike good trade-off between visual quality and computational speed has remained one open challenges in stabilization. Inspired by analogy wobbly frames jigsaw puzzles, we propose an iterative optimization-based learning approach using synthetic datasets for stabilization, which consists two interacting submodules: motion trajectory smoothing full-frame outpainting. First,...

We present MVSGaussian, a new generalizable 3D Gaussian representation approach derived from Multi-View Stereo (MVS) that can efficiently reconstruct unseen scenes. Specifically, 1) we leverage MVS to encode geometry-aware representations and decode them into parameters. 2) To further enhance performance, propose hybrid rendering integrates an efficient volume design for novel view synthesis. 3) support fast fine-tuning specific scenes, introduce multi-view geometric consistent aggregation...

We introduce Probabilistic Coordinate Fields (PCFs), a novel geometric-invariant coordinate representation for image correspondence problems. In contrast to standard Cartesian coordinates, PCFs encode coordinates in correspondence-specific barycentric systems (BCS) with affine invariance. To know \textit{when and where trust} the encoded we implement probabilistic network termed PCF-Net, which parameterizes distribution of fields as Gaussian mixture models. By jointly optimizing their...

Learning-based multi-view stereo (MVS) methods deal with predicting accurate depth maps to achieve an and complete 3D representation. Despite the excellent performance, existing ignore fact that a suitable geometry is also critical in MVS. In this paper, we demonstrate different geometries have significant performance gaps, even using same prediction error. Therefore, introduce ideal composed of Saddle-Shaped Cells, whose predicted map oscillates upward downward around ground-truth surface,...

Regular path query (RPQ) is a basic operation for graph data analysis, and persistent RPQ in streaming graphs new-emerging research topic. In this paper, we propose novel algorithm graphs, named LM-SRPQ. It solves with combination of intermediate result materialization real-time traversal. Compared to prior art, it merges redundant storage computation, achieving higher memory time efficiency. We carry out extensive experiments both real-world synthetic evaluate its performance. Experiment...

Traditional forecasting models mainly utilize historical sales and a few macroeconomic indicators, can no longer reflect the impact of new energy vehicle sales. Based on high-dimensional tensor CNN, this paper proposes new-energy prediction model explores effect data deep learning sales, factors that affect is divided into consumer dimension, dimension social dimension. In each we choose 25 high comprehensive influence factors, then integrate them structure. Through one-dimensional...

With the emergence and popularity of electric vehicles (EVs), number EVs is increasing. However, due to insufficient limited range charging stations, energy problem has held back development EVs. Vehicle-to-Vehicle (V2V) trading seen as a potential solution problem. some problems challenges remain in current V2V scenarios. In order solve selfish behavior among vehicles, we propose an incentive Stackelberg game model facilitate We list functions buyer seller about price they set, then perform...

Excessive carbon emissions will cause the greenhouse effect and global warming, which is not conducive to environmental protection sustainable development. In order realize goal of “carbon peak neutrality” as soon possible, this paper utilizes methodology provided by IPCC measure intensity China’s energy consumption. The classification method emission kernel density function are used explore spatial temporal evolution regional emissions. Based on Log Mean Divided Index (LMDI) method, drivers...

Learning-based multi-view stereo (MVS) method heavily relies on feature matching, which requires distinctive and descriptive representations. An effective solution is to apply non-local aggregation, e.g., Transformer. Albeit useful, these techniques introduce heavy computation overheads for MVS. Each pixel densely attends the whole image. In contrast, we propose constrain augmentation within a pair of lines: each point only corresponding epipolar lines. Our idea takes inspiration from...

This paper proposes a graphene film based flexible millimeter wave bandstop FSS. The has good flexibility and low-density of 1.81 g/cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> . proposed FSS works at 60 GHz with the -10 dB bandwidth 58.33 – 64.83 GHz, which indicates that more than 90% electromagnetic energy is shielded. Furthermore, to expand working bandwidth, double layer sandwich structure (FSS-Foam-FSS) designed. wide 52.78...

In this paper, we present RStab, a novel framework for video stabilization that integrates 3D multi-frame fusion through volume rendering. Departing from conventional methods, introduce perspective to generate stabilized images, addressing the challenge of full-frame generation while preserving structure. The core our approach lies in Stabilized Rendering (SR), rendering module, which extends beyond image by incorporating feature fusion. RStab fusing information space. Specifically, SR...

Generalizable NeRF aims to synthesize novel views for unseen scenes. Common practices involve constructing variance-based cost volumes geometry reconstruction and encoding 3D descriptors decoding views. However, existing methods show limited generalization ability in challenging conditions due inaccurate geometry, sub-optimal descriptors, strategies. We address these issues point by point. First, we find the volume exhibits failure patterns as features of pixels corresponding same can be...

Lensless fiber endomicroscope is an emerging tool for in-vivo microscopic imaging, where quantitative phase imaging (QPI) can be utilized as a label-free method to enhance image contrast. However, existing single-shot reconstruction methods through lensless typically perform well on simple images but struggle with complex structures. Here, we propose speckle-conditioned diffusion model (SpecDiffusion), which reconstructs directly from speckles captured at the detection side of multi-core...

We study the problem of generating intermediate images from image pairs with large motion while maintaining semantic consistency. Due to motion, information may be absent in input images. Existing methods either limit small or focus on topologically similar objects, leading artifacts and inconsistency interpolation results. To overcome this challenge, we delve into pre-trained diffusion models for their capabilities cognition representations, ensuring consistent expression representations...