<p>Aiming at the problem of low accuracy of object detection caused by occlusive and multi-scale clothing, a novel clothing detection algorithm based on improved YOLOv8 is proposed. Firstly, Swin-transformer is adopted to optimize the Backbone network, long-range dependencies information is extracted to improve the detection ability of the model for occlusive clothing. Secondly, the Neck network is reconstructed through the gather-and-distribute mechanism, which gathers and fuses information from all levels and subsequently distributes it to different levels to enhance the information fusion ability. Finally, the receptive field block is introduced to optimize the output of feature fusion to achieve the multi-scale receptive field, improving the object detection ability of the model for multi-scale clothing. The experimental results show that the proposed method achieves excellent performance on the ModaNet dataset, with a significant improvement in the accuracy of detecting occlusive and multi-scale clothing when compared with YOLOv8. Specifically, the accuracy of slight, medium, and heavy occlusive clothing has been improved by 5.2%, 4.6%, and 9.7%, respectively, and the accuracy of small-scale, medium-scale, and large-scale clothing has been improved by 1.4%, 1.2%, and 1.5%, respectively.</p>