Results 1 to 10 of about 22,423,968 (289)

Generalizable Novel-View Synthesis Using a Stereo Camera [PDF]

2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Accepted to CVPR 2024. Project page URL: https://jinwonjoon.github.io/stereonerf/
Haechan Lee, Wonjoon Jin, Seung-Hwan Baek, Sunghyun Cho   +3 more
core   +6 more sources

Improving Structure-From-Motion and Rendering Quality for Thermal Image-Based 3D Gaussian Splatting

IEEE Access
Novel-view synthesis, which generates new views from images captured at multiple viewpoints, is widely applied in fields such as autonomous driving, robotic vision, architecture, and augmented reality.
Jinhoon Wang, Jusang Jeong, Junghyun Oh
doaj   +1 more source

OB3D: A New Dataset for Benchmarking Omnidirectional 3D Reconstruction Using Blender

IEEE Access
Recent advancements in radiance field rendering, exemplified by Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), have significantly progressed 3D modeling and reconstruction.
Shintaro Ito, Natsuki Takama, Toshiki Watanabe, Kent Selwyn The, Koichi Ito, Hwann-Tzong Chen, Takafumi Aoki   +6 more
doaj   +1 more source

GlassGaussian: Extending 3D Gaussian Splatting for Realistic Imperfections and Glass Materials

IEEE Access
3D scene reconstruction methods have made significant advancements in handling reflections. However, these methods are primarily effective on ideal mirror surfaces, which limits their applicability in real-world scenarios where imperfections or materials
Junming Cao, Jiadi Cui, Soren Schwertfeger   +2 more
doaj   +1 more source

FRNeRF: Fusion and regularization fields for dynamic view synthesis

Computational Visual Media
Novel space–time view synthesis for monocular video is a highly challenging task: both static and dynamic objects usually appear in the video, but only a single view of the current scene is available, resulting in inaccurate synthesis results. To address
Xinyi Jing, Tao Yu, Renyuan He, Yu-Kun Lai, Kun Li   +4 more
doaj   +1 more source

Point based representations for novel view synthesis

Représentations basées sur des points pour une nouvelle synthèse de vues L'objectif principal du rendu inverse est de récupérer des informations 3D à partir d'un ensemble d'observations 2D, généralement un ensemble d'images ou de vidéos.
openaire   +3 more sources

Gaussian Splatting: 3D Reconstruction and Novel View Synthesis: A Review

IEEE Access
Image-based 3D reconstruction is a challenging task that involves inferring the 3D shape of an object or scene from a set of input images. Learning-based methods have gained attention for their ability to directly estimate 3D shapes.
Anurag Dalal, Daniel Hagen, Kjell G. Robbersmyr, Kristian Muri Knausgard   +3 more
doaj   +1 more source

3D reconstruction for unconstrained image collections using Gaussian Splatting with foundation model

Geo-spatial Information Science
Achieving high-quality rendering with Gaussian Splatting for unconstrained image collections is critical for advancing 3D reconstruction. Currently, the methods using a single multi-layer perceptron and convolutional neural network to predict distractors
Shuowen Huang, Qingwu Hu, Pengcheng Zhao, Mingyao Ai, Xujie Zhang, Wenwu Ou, Linze Li   +6 more
doaj   +1 more source

NVS-GAN: Benefit of generative adversarial network on novel view synthesis

International Journal of Intelligent Networks
The methodology to generate new views for an object from provided input object view is called Novel View Synthesis (NVS). Humans imagine novel views through prior knowledge gathered through their lifetime.
H.S. Shrisha, V. Anupama
doaj   +1 more source

Single view generalizable 3D reconstruction based on 3D Gaussian splatting

Scientific Reports
3D Gaussian Splatting (3DGS) has become a significant research focus in recent years, particularly for 3D reconstruction and novel view synthesis under non-ideal conditions.
Kun Fang, Qinghui Zhang, Chenxia Wan, Pengtao Lv, Cheng Yuan   +4 more
doaj   +1 more source