Visual Geometry Grounded Transformer
PulseAugur coverage of Visual Geometry Grounded Transformer — every cluster mentioning Visual Geometry Grounded Transformer across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New attention method accelerates 3D scene reconstruction transformers
Researchers have developed a new method called blockwise clustered attention (BC attention) to improve the efficiency of Visual Geometry Grounded Transformers (VGGT), a model used for 3D scene reconstruction. This techn…
-
New method grounds glass surface detection in 3D visual geometry
Researchers have developed a new method for glass surface detection (GSD) by grounding it in 3D visual geometry, moving beyond traditional 2D appearance cues. This approach utilizes a visual geometry grounded transforme…
-
New robot control framework integrates vision and proprioception
Researchers have developed VGGT-DP, a new visuomotor policy framework for robots that integrates geometric priors from a 3D perception model with proprioceptive feedback. This approach aims to improve spatial understand…
-
New multi-view adversarial attack targets 3D vision models
Researchers have developed MVAP-G, a novel method for generating multi-view adversarial perturbations specifically designed to attack the Visual Geometry Grounded Transformer (VGGT). This new technique allows for the cr…
-
New research enhances 3D reconstruction with multi-view geometric priors
A new research paper explores enhancing 3D Gaussian splatting (3DGS) for improved 3D reconstruction quality. The study investigates integrating geometric priors, specifically predicted normal and depth maps, into the 3D…
-
New framework enables calibration-free 3D multi-camera people tracking
Researchers have developed a novel calibration-free framework for 3D multi-camera people tracking in indoor environments. This system integrates multiple deep learning models, including YOLOX for detection, BoT-SORT for…
-
New method enhances 3D scene generation using latent space flow matching
Researchers have developed a new method called Latent Riemannian Flow Matching to improve 3D scene generation using geometric foundation models. This technique operates within the latent space of models like the Visual …
-
EventVGGT framework enhances depth estimation using cross-modal distillation
Researchers have developed EventVGGT, a novel framework for event-based monocular depth estimation that addresses the scarcity of dense depth annotations. This approach leverages cross-modal distillation from Vision Fou…
-
RegimeVGGT accelerates 3D scene reconstruction with layer-wise compression
Researchers have developed RegimeVGGT, a method to improve the efficiency of Visual Geometry Grounded Transformers (VGGTs) for 3D scene reconstruction. Unlike previous methods that applied uniform computation reduction,…
-
RegimeVGGT accelerates 3D scene reconstruction with layer-wise compression
Researchers have developed RegimeVGGT, a novel method to accelerate the Visual Geometry Grounded Transformer (VGGT) for 3D scene reconstruction. By analyzing the layer-specific computational needs, RegimeVGGT applies ta…
-
New MVM-IOD Dataset Evaluates 3D Reconstruction in Industrial Settings
Researchers have introduced the Machine Vision Metrology Industrial Object Dataset (MVM-IOD), a new benchmark designed to evaluate 3D reconstruction and camera pose estimation methods in industrial settings. The dataset…
-
New FGQ method slashes Visual Geometry Transformer model size
Researchers have developed a new post-training quantization method called Fisher-Guided Quantization (FGQ) to reduce the memory and computation overhead of Visual Geometry Grounded Transformers (VGGT). These models, use…
-
Beyond Gaussian Bottlenecks: Topologically Aligned Encoding of Vision-Transformer Feature Spaces
Researchers have developed a new latent learning framework called S$^2$VAE designed to improve the representation of 3D geometry and camera dynamics in visual world models. This approach utilizes a geometry-first perspe…