SGLDet: Spectral-aware Global-Local Dynamic Modeling for Multispectral Aerial Object Detection

Key points

• Multispectral aerial object detection aims to exploit spectral-spatial cues to address object detection challenges in complex aerial scenes.1
• Existing methods mainly focus on spectral-spatial decoupling and object-aware spectral-spatial fusion to improve spectral utilization.1
• However, two critical issues remain underexplored: 1) severe foreground-background imbalance caused by large scene coverage hinders reliable object recognition; 2) aerial objects exhibit substantial local variations in scale, orientation, and structure, which hinder accurate localization.1
• Extensive experiments on MODA, HOD3K, and DroneVehicle demonstrate that SGLDet achieves superior performance with competitive computational cost.1

Sources (1)

• [1]SGLDet: Spectral-aware Global-Local Dynamic Modeling for Multispectral Aerial Object Detection
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 09:06 AM

  To address these issues, we propose SGLDet, a spectral-aware global-local dynamic modeling framework for multispectral aerial object detection, which is motivated by the observation that object semantics are jointly determined by global spectral consistency and scene context, while precise localization requires dynamic feature modulation conditioned on the input spectral and spatial properties.

  Multispectral aerial object detection aims to exploit spectral-spatial cues to address object detection challenges in complex aerial scenes.

Extractive summary: sentences quoted from the sources.