Deep learning-powered object recognition serves as a crucial component within computer vision, finding extensive applications across diverse sectors, including everyday settings and manufacturing industries. This study proposes an innovative object detection algorithm, termed WT-YOLO, which is specifically tailored for the identification of wind turbines. In order to enable the backbone network to effectively extract the appearance features of wind turbines, we propose the ODDS block. Additionally, we incorporate a pyramid squeezing attention module that extracts multi-scale spatial information through a pyramid structure and uses an attention mechanism to focus on effective information in the feature map. During the feature fusion stage, we introduce a progressive feature pyramid network that facilitates direct interaction between adjacent layers, thus mitigating the large semantic gaps between non-adjacent layers. This study assesses the proposed method using a wind turbine dataset and contrasts its performance with YOLOv8 and several SOTA models. The approach attains an mAP50 of 86.6% and an mAP50–95 of 57.2%, highlighting the model’s efficacy in wind turbine detection.