DIL-YOLO: A Content-Adaptive Upsampling and Feature Enhancement Algorithm for Target Detection in UAV Remote Sensing Imagery.

Saved in:

Bibliographic Details
Title:	DIL-YOLO: A Content-Adaptive Upsampling and Feature Enhancement Algorithm for Target Detection in UAV Remote Sensing Imagery.
Authors:	Liu, Siyu¹ 542208907@qq.com, Yin, Hang² 13842205866@163.com
Source:	Engineering Letters. Apr2026, Vol. 34 Issue 4, p1455-1468. 14p.
Subjects:	Object recognition (Computer vision), Drone aircraft, Computer vision, Deep learning, Feature extraction, Image analysis, Interpolation algorithms
Abstract:	Object detection in remote sensing imagery is essential for national security, environmental evaluation, and smart city advancement; however, current methodologies encounter difficulties due to small target dimensions, multi-scale variations, and background clutter. To resolve these challenges, we introduce DIL-YOLO, an innovative detection model featuring significant advancements: the DySample dynamic upsampling operator, which maximizes upsampling efficiency with minimal computational cost; the Inference Feature Enhancement Module (IFM), which bolsters contextual comprehension and enhances small target detection; and the lightweight, structure-aware detection head (LSCDHead), which reduces cross-layer suppression of small targets and enhances weak feature extraction and localization capabilities. Experimental findings indicate that DIL-YOLO surpasses current methodologies, attaining mAP@0.5 enhancements of 3.3% and 2.7% on the VisDrone2019 and DOTA v1.0 datasets, respectively, while employing extremely few parameters. These developments allow DIL-YOLO to effectively identify small objects in intricate remote sensing scenarios, offering significant and dependable assistance for practical applications such as UAV inspection and remote sensing analysis. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Object detection in remote sensing imagery is essential for national security, environmental evaluation, and smart city advancement; however, current methodologies encounter difficulties due to small target dimensions, multi-scale variations, and background clutter. To resolve these challenges, we introduce DIL-YOLO, an innovative detection model featuring significant advancements: the DySample dynamic upsampling operator, which maximizes upsampling efficiency with minimal computational cost; the Inference Feature Enhancement Module (IFM), which bolsters contextual comprehension and enhances small target detection; and the lightweight, structure-aware detection head (LSCDHead), which reduces cross-layer suppression of small targets and enhances weak feature extraction and localization capabilities. Experimental findings indicate that DIL-YOLO surpasses current methodologies, attaining mAP@0.5 enhancements of 3.3% and 2.7% on the VisDrone2019 and DOTA v1.0 datasets, respectively, while employing extremely few parameters. These developments allow DIL-YOLO to effectively identify small objects in intricate remote sensing scenarios, offering significant and dependable assistance for practical applications such as UAV inspection and remote sensing analysis. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X