About
Research & Development
Unlocking Accurate Answers in Visual AI: A Framework for Knowledge-based Question Answering
TLDR: A new training-free framework, KF-VQA, significantly improves Knowledge-based Visual Question Answering (KB-VQA) by addressing knowledge noise. It achieves this through three main components:...
Research & Development
EchoX: Bridging the Semantic-Acoustic Divide in Speech-to-Speech AI
TLDR: EchoX is a new framework for Speech-to-Speech Large Language Models (SLLMs) that addresses the "acoustic-semantic gap" – the disconnect between how speech models...
Research & Development
Enhancing Graph Neural Networks with Flexible Subgraph Pattern Learning
TLDR: MoSE (Mixture of Subgraph Experts) is a novel framework designed to improve Graph Neural Networks (GNNs) by addressing their limitations in capturing complex,...
Research & Development
External Vision System Identifies Unsafe Driving
TLDR: A new vision-based system uses external observation techniques and computer vision (YOLO object detection, custom lane estimation) to classify driver behavior, specifically detecting...
Research & Development
AI-Optimized MRI Protocol Reveals Brain Microstructure Faster
TLDR: Researchers have developed an Explainable AI (XAI) framework to significantly accelerate Diffusion MRI (dMRI) scans for Neurite Exchange Imaging (NEXI) on the Connectome...
Research & Development
ObjectReact: A New Paradigm for Robot Navigation Using Object-Centric Maps
TLDR: ObjectReact is a novel robot navigation system that uses object-relative control instead of traditional image-relative methods. It builds a 3D scene graph of...