Technical University of Denmark
PulseAugur coverage of Technical University of Denmark — every cluster mentioning Technical University of Denmark across labs, papers, and developer communities, ranked by signal.
8 day(s) with sentiment data
-
AI system identifies high-risk pregnancies with notable accuracy
Researchers at the Technical University of Denmark (DTU) have developed an AI system capable of identifying high-risk pregnancies with notable accuracy. This system analyzes various data points to flag potential complic…
-
New research advances 3D Gaussian Splatting for compression and real-world video
Recent research in 3D Gaussian Splatting (3DGS) focuses on improving compression, handling unsynchronized video, and enhancing feed-forward methods. GSCV++ offers lightweight compression for GS sequences by optimizing p…
-
GeoGSplat advances Gaussian Splatting with geometry-grounded approach
Researchers have developed GeoGSplat, a novel approach to Gaussian Splatting that integrates geometric priors to improve novel view synthesis and 3D reconstruction. Unlike previous methods that either align primitives t…
-
New Hexagonal Primitives Enhance 3D Scene Rendering and Reconstruction
Researchers have introduced Honeycomb-Surfels (Hexels), a new method for 3D scene rendering and reconstruction that utilizes compact hexagonal primitives. This approach aims to improve efficiency in memory and computati…
-
CVTSplat method uses adaptive tessellations for faster 3D reconstruction · 2 sources tracked
Researchers have developed CVTSplat, a novel method for multi-view reconstruction that utilizes adaptive centroidal Voronoi tessellations (CVTs). This approach improves upon standard Voronoi tessellations by maintaining…
-
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…
-
Simulation app speeds electric truck battery design and lifetime
A consortium of European universities and industry partners has developed a simulation app to improve the lifetime and robustness of electric truck batteries. Built using COMSOL Multiphysics software, the app, a product…
-
MicroSDF framework enhances 3D reconstruction for mixed-reflectance surfaces
Researchers have developed MicroSDF, a novel neural implicit framework designed to improve 3D reconstruction by effectively handling both reflective and non-reflective surfaces. This new method utilizes microfacet theor…
-
Apple research: Bilingual pre-training boosts LLM performance over hyperparameter tuning
Researchers from Apple Machine Learning Research and the Technical University of Denmark have found that mixing data from a high-resource language, like English, with a low-resource target language, such as Arabic, sign…
-
NEUROTOKEN system enhances auditory attention decoding for noisy environments
Researchers have developed NEUROTOKEN, a novel system for auditory attention decoding (AAD) that aims to solve the cocktail-party problem by identifying which speaker a listener is attending to in noisy environments. Th…
-
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…
-
ZipMVS method significantly reduces memory usage for 3D reconstructions
Researchers have developed ZipMVS, a novel multi-view stereo (MVS) method designed to significantly reduce memory consumption while maintaining high-quality 3D reconstructions. This method employs a unique depth-hypothe…
-
New method uses spherical occupancy profiles for 3D reconstruction and generation
Researchers have developed a novel method for 3D reconstruction and generation using spherical occupancy profiles, which represent ray-wise occupancy probabilities. This approach unifies discriminative and generative ta…
-
New framework uses epipolar geometry for sparse-view 3D reconstruction
Researchers have developed EpiS, a novel neural surface reconstruction framework designed to overcome challenges in creating accurate 3D models from sparse multi-view images. Unlike previous methods that rely on simplif…
-
New method enables structured asset creation from 3D Gaussian Splatting scenes
Researchers have developed Manifold-GS, a novel method for creating structured assets from 3D Gaussian Splatting (3DGS) scenes. Unlike traditional 3DGS, Manifold-GS separates appearance opacity from geometric mass and r…
-
Danish university plans to phase out pen-and-paper exams, sparking educator protests
DTU, a Danish technical university, plans to eliminate pen-and-paper exams starting in the 2027/2028 academic year. This decision has sparked a protest from 35 educators across DTU Physics and DTU Nanolab, who argue tha…
-
G2SR method offers fast, memory-efficient 3D surface reconstruction
Researchers have developed G2SR, a novel method for fast and memory-efficient 3D surface reconstruction using Gaussian-based techniques. Unlike existing end-to-end neural network approaches that demand significant compu…
-
New Stochastic Signed Distance Processes enhance surface reconstruction
Researchers have introduced Stochastic Signed Distance Processes (SSDP), a novel approach to multi-view surface reconstruction that models the Signed Distance Field (SDF) along each ray as a stochastic process. This pro…
-
New framework TRUST-TSE tackles shortcut learning in EEG-guided speech extraction
Researchers have developed a new framework called TRUST-TSE to improve the reliability of EEG-guided target speech extraction. This method addresses the issue of shortcut learning, where models perform well within a sin…
-
New framework TRUST-TSE tackles shortcut learning in EEG-guided speech extraction
Researchers have developed a new framework called TRUST-TSE to address shortcut learning in EEG-guided target speech extraction. This method aims to improve the generalization of neuro-steered hearing technologies by pr…