This paper revisits Deep Geometric Moments (DGM), a framework for learning shape-aware, geometry-aligned visual representations, by swapping the original ResNet backbone for ConvNeXt, a modern high-performing ConvNet. Using ConvNeXt feature maps as the input “stem” for DGM, the authors show consistent ImageNet-1K accuracy gains over a ResNet34-DGM baseline (up to ~+2.4% depending on ConvNeXt size) while preserving and strengthening DGM’s shape bias.
A comprehensive review on AI-powered wearable biosensors, highlighting how machine learning and edge AI enable real-time health monitoring and personalized care, including digital twins, LLMs, and challenges in privacy, scalability, and clinical integration.
Proposed a novel framework for generating CFs using large language models (LLMs), with a focus on structured sensor-derived datasets in health and physiological monitoring