The study presents a process of digital simulation that aims to investigate the legibility of multiple spaces in a complex architecture through architectural survey, virtualreconstruction and3dvisualization in imme...
详细信息
The study presents a process of digital simulation that aims to investigate the legibility of multiple spaces in a complex architecture through architectural survey, virtualreconstruction and3dvisualization in immersive environment. The collaboration between two research institutions, one Italian and the other Chinese, developed a reconstruction of a building in the campus of the Tsinghua University of Beijing using digital tools, in order to understand the behaviours during the fruition of that space. digital simulation in the visibility theme guarantees the possibility of testing multiple configurations and showing the impacts of the different environmental hypothesis.
The data presented by the Ministry of Public Health of Russia over the past 10 years show that the incidence of the malignant tumours in the population has been increasing by 1,5% annually. Unfortunately, more th...
详细信息
Reconstructing both objects and hands in 3d from a single RGB image is complex. Existing methods rely on manually defined hand-object constraints in Euclidean space, leading to suboptimal feature learning. Compared wi...
Reconstructing both objects and hands in 3d from a single RGB image is complex. Existing methods rely on manually defined hand-object constraints in Euclidean space, leading to suboptimal feature learning. Compared with Euclidean space, hyperbolic space better preserves the geometric properties of meshes thanks to its exponentially-growing space distance, which amplifies the differences between the features based on similarity. In this work, we propose the first precise hand-object reconstruction method in hyperbolic space, namely dynamic Hyperbolic Attention Network (dHANet), which leverages intrinsic properties of hyperbolic space to learn representative features. Our method that projects mesh and image features into a unified hyperbolic space includes two modules, i.e. dynamic hyperbolic graph convolution and image-attention hyperbolic graph convolution. With these two modules, our method learns mesh features with rich geometry-image multi-modal information and models better hand-object interaction. Our method provides a promising alternative for fine hand-object reconstruction in hyperbolic space. Extensive experiments on three public datasets demonstrate that our method outperforms most state-of-the-art methods.
With the intention of providing a general idea of the process and computational problems to recover a 3d scene from a point cloud, this paper presents a state-of-the-art review on point cloud generation from images fo...
详细信息
In 3d surface reconstruction, the sampling and quantization in the process of digitization limits the surface details obtained by the visual sensor. Obtaining a fine-grained3d surface at lower cost remains a difficul...
详细信息
Small unmanned aerial vehicles (UAVs) can use their flexible maneuverability to effectively collect images of urban buildings, and use the current multi-view stereo method to generate advanced3d urban models. But the...
详细信息
The development of a digital management system to support historical building assets39; management process based on public participation is the latest innovation in preservation, conservation, and restoration activi...
详细信息
deep learning applied to the reconstruction of 3d shapes has seen growing interest. A popular approach to 3dreconstruction and generation in recent years has been the CNN encoder-decoder model usually applied in voxe...
详细信息
In computer vision, object reconstruction is the task of inferring the 3d shape of an object based on a single or multiple 2d images. For such purpose, most common frameworks use voxel grids and point clouds. However,...
详细信息
ISBN:
(纸本)9781510635241
In computer vision, object reconstruction is the task of inferring the 3d shape of an object based on a single or multiple 2d images. For such purpose, most common frameworks use voxel grids and point clouds. However, both of these approaches have strong limitations. On one hand, the computational cost of using voxels grows cubically as the resolution of the voxels increases. Therefore, 3d object reconstructions are usually set to low resolution. On the other hand, point clouds are unstructured in nature and the proper definition of surfaces and contours is complex. In this study, 3d object reconstruction is carried out applying free-form deformations on pre-existent 3d meshes, through two basic learning processes: template selection and template deformation. From this approach, it is possible to generate high-quality 3d object reconstructions with a lower computational cost. Concretely, two novel lightweight CNNs models are developed and tested: a multi-target learner (Model A) anddepth information learner (Model B). According to the results, the performance of the multi-target learner regarding the template selection was around three times better (lower error) than in the baseline architecture, which improved the quality of the 3dreconstructions, whereas the depth-information learner showed promising results in the reconstruction of objects with complex geometry. The inherent issue of using chamfer distance as a loss measure is also examined.
The proceedings contain 84 papers. The topics discussed include: sign-aware perturbations regression;Frank-Wolfe algorithm for learning SVM-type multi-category classifiers;reconstruction-based anomaly detection with c...
ISBN:
(纸本)9781611976700
The proceedings contain 84 papers. The topics discussed include: sign-aware perturbations regression;Frank-Wolfe algorithm for learning SVM-type multi-category classifiers;reconstruction-based anomaly detection with completely random forest;do winning tickets exist before dNN training?;mining easily understandable models from complex event logs;deep neural network for 3d surface segmentation based on contour tree;learning time-series shapelets via supervised feature selection;fair classification under strict unawareness;provable distributed stochastic gradient descent with delayed updates;disentangleddynamic graph deep generation;robust dual recurrent neural networks for financial time series;and a fine-grained graph-based spatiotemporal network for bike flow prediction in bike-sharing systems.
暂无评论