Neural Radiance Fields
PulseAugur coverage of Neural Radiance Fields — every cluster mentioning Neural Radiance Fields across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New 3D Gaussian method enables fast relighting of human portraits
Researchers have developed GSHeadRelight, a new method for synthesizing and relighting 3D human heads from in-the-wild images. This approach utilizes 3D Gaussians and a neural radiance transfer representation to achieve…
-
New method aligns neural network weights, reducing barriers in NeRF models
Researchers have developed a new method called Progressive Permutation Alignment with Temperature-Controlled Spectral Sharpening to address the challenge of permutation symmetry in implicit neural representations (INRs)…
-
DriftNeRF improves sparse-view Neural Radiance Fields by penalizing density disagreement
Researchers have developed DriftNeRF, a novel regularizer designed to improve the performance of Neural Radiance Fields (NeRF) when only a sparse set of input views is available. The method addresses a core limitation i…
-
New benchmark evaluates camera pose estimation for 3D scene synthesis
Researchers have developed a benchmark to evaluate camera pose estimation strategies for novel view synthesis, a process crucial for creating 3D scene representations. The study found that traditional Structure-from-Mot…
-
VisibilitySplat enhances 3D Gaussian Splatting for inverse rendering
Researchers have developed VisibilitySplat, a novel method that enhances 3D Gaussian Splatting (3DGS) for inverse rendering tasks. This approach integrates explicit visibility modeling into the standard 3DGS pipeline, e…
-
RouteBridge framework enhances NeRF and 3D Gaussian Splatting distillation
Researchers have developed RouteBridge, a novel bidirectional framework designed to improve the distillation process between Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting (3DGS). This system adaptively select…
-
New TR-NeRF method dramatically boosts 3D human body reconstruction efficiency
Researchers have developed Temporal Residual Neural Radiance Fields (TR-NeRF) to improve the efficiency and quality of 3D human body reconstruction from monocular video. This new method addresses the limitations of trad…
-
AccidentSim generates realistic vehicle collision videos from accident reports
Researchers have developed AccidentSim, a new framework designed to generate realistic vehicle collision videos for autonomous driving research. This system extracts physical clues and contextual information from real-w…
-
Quantum-classical model 3D-QISR advances novel view synthesis
Researchers have introduced 3D Quantum Implicit Scene Representation (3D-QISR), a novel hybrid quantum-classical model for novel-view synthesis. This approach integrates parameterised quantum circuits into the Neural Ra…
-
Small UAVs achieve high-fidelity 3D object reconstruction
Researchers have developed a novel system enabling small, sub-100-gram unmanned aerial vehicles (UAVs) to autonomously perform high-fidelity 3D reconstruction of static objects. The system features a closed-loop active …
-
New AI methods enhance 3D scene editing with multi-view consistency and LLM integration · 6 sources tracked
Researchers are developing new methods to improve the consistency and accuracy of 3D scene editing using AI. One approach, "3D-Consistent Multi-View Editing by Correspondence Guidance," focuses on ensuring that correspo…
-
New PIVOT dataset evaluates 3D reconstruction under real-world conditions
Researchers have introduced PIVOT, a new dataset and evaluation framework designed to assess the performance of 3D reconstruction methods like Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting (3DGS) in more real…
-
New filter tackles aliasing in 4D Gaussian representations for dynamic scenes
Researchers have developed a new motion-aware filtering technique to reduce aliasing artifacts in 4D Gaussian representations, which are used for synthesizing novel views of dynamic scenes in AR/VR applications. Existin…
-
FixAnything model refines 3D rendering artifacts using video generative priors
Researchers have developed FixAnything, a novel model designed to refine 3D renderings and fix artifacts across various representations like Gaussian Splatting and NeRF. The model repurposes a pre-trained video generati…
-
New NeRF-based SLAM system tackles large-scale scene reconstruction
Researchers have developed a new approach to Neural Radiance Fields (NeRF)-based Simultaneous Localization and Mapping (SLAM) designed for large-scale environments. This system utilizes a multi-submap architecture and a…
-
New ERF-GS framework reconstructs fast motion using event-RGB fusion
Researchers have developed ERF-GS, a novel framework for reconstructing fast-moving objects in 3D scenes. This method integrates event data from high-frame-rate sensors into the Gaussian splatting pipeline, improving ac…
-
New Convolutional Neural Shading Method Enhances 3D Reconstruction
Researchers have introduced Convolutional Neural Shading (CNS), a new method for generating high-quality 3D reconstructions from multiple images. Unlike previous neural rendering techniques that rely on limited geometri…
-
Survey paper details Neural Radiance Fields advancements and challenges
A new survey paper published on arXiv details the advancements and challenges in Neural Radiance Fields (NeRFs). The paper, authored by Wenhui Xiao, provides a comprehensive review of theoretical innovations, alternativ…
-
New ICEF framework enhances few-shot facial expression synthesis
Researchers have developed Identity-Consistent Expression Fields (ICEF), a new framework for synthesizing facial expressions from a limited set of images. This method disentangles identity-specific features from express…
-
BathyFacto: New NeRF Model Accurately Maps Underwater Terrain
Researchers have developed BathyFacto, a novel extension of Neural Radiance Fields (NeRF) designed to accurately map underwater terrain. This new method addresses the challenges of refraction at the air-water interface,…