Article Info
Author(s) Uteuliev N.U., Dauletnazarov J.I.
DOI DOI 10.24412/2181-1431-2025-4-83-88
Abstract
This research utilizes a proprietary dataset consisting of images captured from cotton fields
in the Republic of Karakalpakstan. A new annotated image dataset from cotton fields in the Nukus district
was developed to compare the performance of YOLOv8 and YOLOv9 models in multi-class weed
detection. YOLOv9c has outperformed the competing model YOLOv8m with respect to overall accuracy
(mAP@0.5 = 0.865). YOLOv9c is therefore more appropriate for ongoing precise weed management
during real-time applications. YOLOv8m, however, may still be deployed on energy-efficient hardware