.png){kind=icon}
.png){kind=icon}
Return
ADCF-YOLO: Adaptive Dynamic Context Fusion YOLO for Remote Sensing Object Detection
.png){kind=icon}
PRE
AI
J
H
H
DOI：10.1109/lgrs.2026.3711846.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Small targets in remote sensing images exhibit significant scale variations and weak texture features, making accurate detection challenging. To address these issues, this letter proposes adaptive dynamic context fusion YOLO (ADCF-YOLO). The model comprises four collaboratively designed components: EBFPN builds a bidirectional feature pyramid with a dedicated high-resolution detection layer and cross-layer connections to preserve fine-grained spatial cues; MBEM reinforces local edge and texture responses in high-resolution backbone layers; TCAM captures long-range global context through spatial-channel collaborative attention in the neck; and AMDA-Head integrates multiscale features via complementary soft and hard attention at the detection boundary. Together, they form a systematic local-to-global feature reinforcement chain targeting the two core bottlenecks of remote sensing small-target detection. Experimental results show that ADCF-YOLO achieves 4% and 7.2% improvements in mAP50 and mAP50:95 over YOLO11n on DIOR, with 8.1% and 2.4% mAP50 gains on VEDAI and SIMD.
Keywords:
Adaptive mechanism
feature fusion network
global information weighting
remote sensing images
Journal
I
IF:
4.4
Papers:
486
Citations:
0
Organization
Cited Papers