NYU-Depth V2
PulseAugur coverage of NYU-Depth V2 — every cluster mentioning NYU-Depth V2 across labs, papers, and developer communities, ranked by signal.
-
GeoBridge framework evolves transferable decoder topologies for geometric vision
Researchers have introduced GeoBridge, an evolutionary framework designed to discover transferable decoder topologies for geometric vision tasks. This system optimizes internal computations within decoders, allowing the…
-
Bio-inspired Monocentric Imaging enhances 3D sensing with novel optical design
Researchers have developed a novel Bio-inspired Monocentric Imaging (BMI) framework to improve large-field-of-view 3D sensing for autonomous systems. This framework integrates a monocentric optical design with a spheric…
-
New datasets and lightweight models advance monocular depth estimation
Researchers are developing new methods and datasets for monocular depth estimation, a technique crucial for applications like augmented and virtual reality. New datasets such as MODEST are being created to provide high-…
-
New Condition Dropout method boosts RGB-D segmentation robustness
Researchers have developed a new method called Condition Dropout (ConD) to improve the robustness of RGB-D semantic segmentation models. These models typically require both RGB and depth data, but practical sensor failu…
-
New UniM2 framework enables unsupervised multimodal semantic segmentation
Researchers have introduced UniM2, a novel framework designed for Unsupervised Multimodal Semantic Segmentation (UMSS). This approach aims to effectively leverage complementary sensor information without requiring any l…
-
Vision models fail to verify physical causality, new research finds
A new research paper titled "Geometric Collapse: When Vision Models Fail to Verify Physical Causality" introduces a controlled counterfactual called Scrambled Edges. This method injects edge-like cues into visual data w…
-
Vision Transformers learn spatial hierarchy mirroring primate visual cortex
Researchers have investigated how Vision Transformers (ViTs) encode spatial information without explicit spatial supervision during pretraining. By probing a ViT-B/16 model, they found that boundary structure is decodab…
-
Monocular Depth Estimation via Neural Network with Learnable Algebraic Group and Ring Structures
Researchers have developed LAGRNet, a new framework for monocular depth estimation that incorporates algebraic geometry principles. Unlike previous methods that treat depth estimation as a generic regression problem, LA…