The field of 3D reconstruction and scene understanding is rapidly advancing, with a focus on improving the accuracy and efficiency of various methods. Recent developments have led to the proposal of novel approaches for textured mesh quality assessment, Gaussian Splatting, and depth estimation. These methods have shown promising results in terms of robustness and generalization capability. Notably, the integration of monocular priors and geometry-aware techniques has improved the performance of stereo matching and 3D reconstruction algorithms. Furthermore, the development of new datasets and evaluation metrics has facilitated the comparison of different methods and driven the progress of the field. Overall, the current trends in 3D reconstruction and scene understanding are moving towards more accurate, efficient, and generalized models. Some noteworthy papers in this regard include EA-3DGS, which proposes an efficient and adaptive 3D Gaussians method for outdoor scenes, and MonoSplat, which introduces a generalizable 3D Gaussian Splatting framework from monocular depth foundation models. Additionally, the paper on Synthetic Enclosed Echoes presents a novel dataset for underwater 3D reconstruction, and the work on SHaDe introduces a compact and consistent dynamic 3D reconstruction method via tri-plane deformation and latent diffusion.
Advances in 3D Reconstruction and Scene Understanding
Sources
Textured mesh Quality Assessment using Geometry and Color Field Similarity
EA-3DGS: Efficient and Adaptive 3D Gaussians with Highly Enhanced Quality for outdoor scenes
Multi-view dense image matching with similarity learning and geometry priors
Multi-Label Stereo Matching for Transparent Scene Depth Estimation
M3Depth: Wavelet-Enhanced Depth Estimation on Mars via Mutual Boosting of Dual-Modal Data
Towards Generating Realistic Underwater Images
Diving into the Fusion of Monocular Priors for Generalized Stereo Matching
SparC: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling
Personalize Your Gaussian: Consistent 3D Scene Personalization from a Single Image
3D Reconstruction from Sketches
MonoSplat: Generalizable 3D Gaussian Splatting from Monocular Depth Foundation Models
GT^2-GS: Geometry-aware Texture Transfer for Gaussian Splatting
R3GS: Gaussian Splatting for Robust Reconstruction and Relocalization in Unconstrained Image Collections
Synthetic Enclosed Echoes: A New Dataset to Mitigate the Gap Between Simulated and Real-World Sonar Data
RUSplatting: Robust 3D Gaussian Splatting for Sparse-View Underwater Scene Reconstruction
Constructing a 3D Town from a Single Image
SHaDe: Compact and Consistent Dynamic 3D Reconstruction via Tri-Plane Deformation and Latent Diffusion
Masked Conditioning for Deep Generative Models
Seeing through Satellite Images at Street Views