Results 1 to 10 of about 22,423,978 (289)

Depth-Supervised and Curvature-Optimized 3D Gaussian Splatting for Scene Reconstruction

IEEE Access
Recently, 3DGS has achieved revolutionary advancements in dynamic rendering and novel view synthesis tasks. The Adaptive Density Control (ADC) algorithm plays a pivotal role in 3DGS, utilizing an image gradient-based average gradient threshold to ...
Antong Li, Lutao Wang, Ziwei Wang
doaj   +1 more source

iNVS: Repurposing Diffusion Inpainters for Novel View Synthesis

SIGGRAPH Asia 2023 Conference Papers, 2023
Accepted to SIGGRAPH Asia, 2023 (Conference Papers)
Yash Kant, Aliaksandr Siarohin, Michael Vasilkovsky, Riza Alp Güler, Jian Ren 0005, Sergey Tulyakov, Igor Gilitschenski   +6 more
openaire   +4 more sources

Cross-View Geo-Localization via 3D Gaussian Splatting-Based Novel View Synthesis

Remote Sensing
Cross-view geo-localization allows an agent to determine its own position by retrieving the same scene from images taken from dramatically different perspectives.
Xiaokun Ding, Xuanyu Zhang, Shangzhen Song, Bo Li, Le Hui, Yuchao Dai   +5 more
doaj   +1 more source

IVAH: Invisible-Region Augmented Human Rendering From Monocular Images

IEEE Access
Most neural radiance field (NeRF)-based human rendering methods rely on per-subject optimization or multi-view data input to reconstruct the 3D radiance field of human subjects.
Do Yeop Kim, Ju Yong Chang
doaj   +1 more source

Novel View Synthesis of Humans Using Differentiable Rendering

IEEE Transactions on Biometrics, Behavior, and Identity Science, 2023
We present a new approach for synthesizing novel views of people in new poses. Our novel differentiable renderer enables the synthesis of highly realistic images from any viewpoint. Rather than operating over mesh-based structures, our renderer makes use of diffuse Gaussian primitives that directly represent the underlying skeletal structure of a human.
Guillaume Rochette, Chris Russell 0001, Richard Bowden   +2 more
openaire   +4 more sources

Depth-Gradient-Guided Decoupled Optimization in 3D Gaussian Splatting for Sparse-View Reconstruction

Applied Sciences
3D Gaussian Splatting (3DGS) achieves high-quality novel-view synthesis under dense inputs but suffers from severe overfitting and geometric artifacts in sparse-view settings.
Xiang Lian, Zongji Wang, Hong Wang
doaj   +1 more source

Novel View Synthesis as Video Completion

CoRR
Project page: https://frame-crafter.github.io/
Qi Wu, Khiem Vuong, Minsik Jeon, Srinivasa G. Narasimhan, Deva Ramanan   +4 more
openaire   +3 more sources

Novel view synthesis in embedded virtual reality devices

Electronic Imaging, 2022
info:eu-repo/semantics ...
Laurie Van Bogaert, Daniele Bonatto, Sarah Fernades Pinto Fachada, Gauthier Lafruit   +3 more
openaire   +2 more sources

TGA-GS: Thermal Geometrically Accurate Gaussian Splatting

Applied Sciences
Novel view synthesis and 3D reconstruction have been extensively studied. Three-dimensional Gaussian Splatting (3DGS) has gained popularity due to its rapid training and real-time rendering capabilities.
Chen Zou, Qingsen Ma, Jia Wang, Rongfeng Lu, Ming Lu, Zhaowei Qu   +5 more
doaj   +1 more source

V4D: Voxel for 4D Novel View Synthesis

IEEE Transactions on Visualization and Computer Graphics
Neural radiance fields have made a remarkable breakthrough in the novel view synthesis task at the 3D static scene. However, for the 4D circumstance (e.g., dynamic scene), the performance of the existing method is still limited by the capacity of the neural network, typically in a multilayer perceptron network (MLP).
Wanshui Gan, Hongbin Xu, Yi Huang 0035, Shifeng Chen, Naoto Yokoya   +4 more
openaire   +4 more sources