Scholarlink

<p>Addressing the challenges of small object detection in UAV-captured infrared images, this study introduces DiSTA-DETR, an enhanced detection model based on RT-DETR-R18. The model incorporates an enhanced dilated transformer attention (EDTA) module for expanded receptive fields and robust small object representation, an enhanced guided context aggregation (EGCA) module for multi-scale feature extraction, and a hybrid sparse enhancement (HSE) module for global modeling and fine-detail preservation. Experiments on the HIT-UAV dataset demonstrate significant improvements, with a 5.9% increase in mAP50 and a 3.7% increase in mAP50:95, confirming the model's effectiveness in complex scenarios. The code and associated datasets are available at <a href="https://github.com/88kucheng/DiSTA-DETR">https://github.com/88kucheng/DiSTA-DETR</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced infrared small object detection in UAV imagery: DiSTA-DETR with dilated-sparse thermal attention

  • 期刊论文

  • 发表日期：

    2026年5月10日

查看全文
  • Yongcheng Wang,
  • Yao Rao,
  • Gangping Zhang,
  • Chaofeng Li

摘要

Addressing the challenges of small object detection in UAV-captured infrared images, this study introduces DiSTA-DETR, an enhanced detection model based on RT-DETR-R18. The model incorporates an enhanced dilated transformer attention (EDTA) module for expanded receptive fields and robust small object representation, an enhanced guided context aggregation (EGCA) module for multi-scale feature extraction, and a hybrid sparse enhancement (HSE) module for global modeling and fine-detail preservation. Experiments on the HIT-UAV dataset demonstrate significant improvements, with a 5.9% increase in mAP50 and a 3.7% increase in mAP50:95, confirming the model's effectiveness in complex scenarios. The code and associated datasets are available at https://github.com/88kucheng/DiSTA-DETR.