LeC$^2$O-NeRF: Learning Continuous and Compact Large-Scale Occupancy for Urban Scenes
This paper introduces LeC$^2$O-NeRF, a novel method for learning a continuous and compact occupancy representation for large-scale urban scenes in Neural Radiance Fields (NeRFs) to significantly improve the efficiency of NeRF training without sacrificing accuracy.
Planar Reflection-Aware Neural Radiance Fields for Improved Novel View Synthesis
This paper introduces a novel method for enhancing Neural Radiance Fields (NeRF) to accurately model and render planar reflections, improving the realism and accuracy of scene reconstruction for novel view synthesis.
ProvNeRF: Enhancing Neural Radiance Fields with a Stochastic Field for Modeling Point Visibility
ProvNeRF improves the accuracy and uncertainty estimation of Neural Radiance Fields (NeRFs) by explicitly modeling the probability distribution of camera positions from which each 3D point is visible, thereby enhancing scene reconstruction and enabling more reliable uncertainty quantification in challenging sparse view scenarios.
Gumbel-NeRF: Enhancing Neural Radiance Fields for Synthesizing Novel Views of Unseen Objects Using a Mixture-of-Expert Approach
Gumbel-NeRF improves upon existing Neural Radiance Field methods by introducing a novel expert selection mechanism and training strategy, enabling the synthesis of high-quality novel views of unseen objects from limited input data.
Neural Radiance Field Image Enhancement via End-to-End Optimized Sampling Point Placement
Optimizing sampling point placement within a Neural Radiance Field framework, using an MLP-Mixer inspired architecture, reduces rendering artifacts and improves the quality of novel viewpoint image generation.
Depth-Supervised Neural Radiance Fields: Leveraging Depth Information for Improved View Synthesis and Faster Training
Incorporating readily available depth information as supervision significantly improves the performance of Neural Radiance Fields (NeRF) for view synthesis, particularly in scenarios with limited training views, by accelerating training and enhancing the accuracy of rendered geometry.
TD-NeRF: Optimizing Neural Radiance Fields and Camera Poses Jointly Using Monocular Depth Priors
TD-NeRF leverages readily available monocular depth priors to simultaneously optimize camera poses and neural radiance fields, achieving superior performance in novel view synthesis and pose estimation, particularly in challenging scenarios with large motion changes.
Deformable Neural Radiance Fields Using Recursively Subdivided Tetrahedra for Efficient 3D Object Manipulation and Animation
DeformRF is a novel method that enhances Neural Radiance Fields (NeRFs) by integrating deformable tetrahedral meshes, enabling efficient and realistic 3D object manipulation and animation while maintaining high-quality rendering.
H2O-SDF: Two-Phase Learning for 3D Indoor Reconstruction
새로운 두 단계 학습 방법을 소개하여 3D 실내 재구성을 개선합니다.
NeWRF: Ein tiefes Lernrahmenwerk für die Rekonstruktion des drahtlosen Strahlungsfelds und die Kanalvorhersage
Ein tiefes Lernrahmenwerk, NeWRF, ermöglicht präzise Vorhersagen drahtloser Kanäle mit geringer Messdichte.
© 2024 by Linnk AI