HiCoM: A Novel Framework for Efficient Online Reconstruction of Streamable Dynamic Scenes Using 3D Gaussian Splatting
HiCoM is a novel framework for efficient online reconstruction of streamable dynamic scenes that leverages a hierarchical coherent motion mechanism and continual refinement to achieve faster training, reduced storage, and competitive rendering quality compared to state-of-the-art methods.
Real-Time 4D Reconstruction of Dynamic Endoscopic Scenes Using 4D Gaussian Splatting for Enhanced Surgical Precision
This research paper introduces ST-Endo4DGS, a novel framework that leverages 4D Gaussian Splatting to achieve real-time, high-fidelity reconstruction of dynamic endoscopic scenes, enhancing surgical precision in minimally invasive procedures.
ReMatching: Improving Dynamic Scene Reconstruction with Deformation Priors
The ReMatching framework improves the generalization quality of dynamic scene reconstruction models by incorporating deformation priors, specifically velocity-field-based priors, through a novel flow-matching loss.
Direction-aware Representations for Dynamic Scene Reconstruction using NeRF and Gaussian Splatting
DaRePlane, a novel direction-aware representation based on the dual-tree complex wavelet transform (DTCWT), enhances dynamic scene reconstruction in both NeRF and Gaussian Splatting frameworks by overcoming limitations of traditional wavelet representations.
Free-DyGS: A Novel Method for Camera-Pose-Free 4D Scene Reconstruction of Dynamic Surgical Videos Using Gaussian Splatting
This paper introduces Free-DyGS, a novel camera-pose-free framework for reconstructing dynamic surgical scenes from endoscopic videos using 3D Gaussian splatting, addressing challenges in accuracy, camera positioning, dynamic scenes, and reconstruction speed.
MonST3R: A Geometry-First Approach for Estimating Dynamic Scene Geometry from Videos
Directly estimating per-timestep geometry as pointmaps, trained with a specific focus on dynamic scenes, offers a robust and efficient method for reconstructing dynamic scenes from videos.
High-Fidelity Dynamic Mesh Reconstruction from Monocular Videos
DG-Mesh, a framework that reconstructs high-quality, time-consistent meshes from monocular videos, leveraging 3D Gaussian Splatting and a novel Gaussian-Mesh Anchoring procedure.
DrivingGaussian: Composite Gaussian Splatting for Dynamic Autonomous Driving Scenes
DrivingGaussian introduces Composite Gaussian Splatting for high-quality dynamic scene reconstruction in autonomous driving.
DrivingGaussian: Composite Gaussian Splatting for Dynamic Autonomous Driving Scenes
DrivingGaussian introduces Composite Gaussian Splatting to efficiently represent dynamic autonomous driving scenes, outperforming existing methods and enabling high-quality synthesis of surrounding views.
Motion-aware 3D Gaussian Splatting for Efficient Dynamic Scene Reconstruction
Enhancing dynamic 3D Gaussian Splatting with motion cues from optical flow for efficient scene reconstruction.
© 2024 by Linnk AI